diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..2a2119c2ae4c4d652062bbd09e99297394b111ea
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/README.md
@@ -0,0 +1,58 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: transformers
+model_name: Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1
+tags:
+- generated_from_trainer
+- trl
+- sft
+licence: license
+---
+
+# Model Card for Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1
+
+This model is a fine-tuned version of [Qwen/Qwen3-14B-Base](https://huggingface.co/Qwen/Qwen3-14B-Base).
+It has been trained using [TRL](https://github.com/huggingface/trl).
+
+## Quick start
+
+```python
+from transformers import pipeline
+
+question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
+generator = pipeline("text-generation", model="None", device="cuda")
+output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
+print(output["generated_text"])
+```
+
+## Training procedure
+
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/katriin-kukk/Cross_lingual_morphological_generalization/runs/u51ovefm) 
+
+
+
+This model was trained with SFT.
+
+### Framework versions
+
+- TRL: 0.29.0
+- Transformers: 5.5.4
+- Pytorch: 2.10.0
+- Datasets: 4.6.1
+- Tokenizers: 0.22.2
+
+## Citations
+
+
+
+Cite TRL as:
+    
+```bibtex
+@software{vonwerra2020trl,
+  title   = {{TRL: Transformers Reinforcement Learning}},
+  author  = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin},
+  license = {Apache-2.0},
+  url     = {https://github.com/huggingface/trl},
+  year    = {2020}
+}
+```
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..e4b899469577e05af5031b26ff3af4ff3a41bef0
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/README.md
@@ -0,0 +1,58 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: transformers
+model_name: Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2
+tags:
+- generated_from_trainer
+- trl
+- sft
+licence: license
+---
+
+# Model Card for Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2
+
+This model is a fine-tuned version of [Qwen/Qwen3-14B-Base](https://huggingface.co/Qwen/Qwen3-14B-Base).
+It has been trained using [TRL](https://github.com/huggingface/trl).
+
+## Quick start
+
+```python
+from transformers import pipeline
+
+question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
+generator = pipeline("text-generation", model="None", device="cuda")
+output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
+print(output["generated_text"])
+```
+
+## Training procedure
+
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/katriin-kukk/Cross_lingual_morphological_generalization/runs/q6nyoshd) 
+
+
+
+This model was trained with SFT.
+
+### Framework versions
+
+- TRL: 0.29.0
+- Transformers: 5.5.4
+- Pytorch: 2.10.0
+- Datasets: 4.6.1
+- Tokenizers: 0.22.2
+
+## Citations
+
+
+
+Cite TRL as:
+    
+```bibtex
+@software{vonwerra2020trl,
+  title   = {{TRL: Transformers Reinforcement Learning}},
+  author  = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin},
+  license = {Apache-2.0},
+  url     = {https://github.com/huggingface/trl},
+  year    = {2020}
+}
+```
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..5dcee5333d469fb53410e46d1ad03aba1acdd9fb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/trainer_state.json
@@ -0,0 +1,139 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.24125452352231605,
+  "eval_steps": 20,
+  "global_step": 100,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.0561259958272e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..8a42c886c671421eb0616f8aa3810f2e8675962f
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/trainer_state.json
@@ -0,0 +1,1084 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.4101326899879374,
+  "eval_steps": 20,
+  "global_step": 1000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.0729150094336e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb1d776a35ca219c34fe19e0b83270b6e78a3ecc
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/trainer_state.json
@@ -0,0 +1,1105 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.4583835946924006,
+  "eval_steps": 20,
+  "global_step": 1020,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.17605631420416e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..3424abaefaa32182f7253121f1c7203b530b067d
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/trainer_state.json
@@ -0,0 +1,1126 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.5066344993968634,
+  "eval_steps": 20,
+  "global_step": 1040,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.27477297872896e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..bb23a6318878332851b486b116475f10f6955772
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/trainer_state.json
@@ -0,0 +1,1147 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.554885404101327,
+  "eval_steps": 20,
+  "global_step": 1060,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.370774678528e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..266f3a155d3fa35243664affc0f53f0d50547da6
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/trainer_state.json
@@ -0,0 +1,1168 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.60313630880579,
+  "eval_steps": 20,
+  "global_step": 1080,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.47206953373696e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..82b8461fae59c4c1d25b26d55d0d24742aeed947
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/trainer_state.json
@@ -0,0 +1,1189 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.651387213510253,
+  "eval_steps": 20,
+  "global_step": 1100,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.57585367650304e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..f7b2dd1ee40f08c5e9366aa2c12ffadf35fb49d0
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/trainer_state.json
@@ -0,0 +1,1210 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.6996381182147164,
+  "eval_steps": 20,
+  "global_step": 1120,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.68034904428544e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..281e3ae2be6d3cc797a50ab3a9cd053d65d77680
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/trainer_state.json
@@ -0,0 +1,1231 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.7478890229191797,
+  "eval_steps": 20,
+  "global_step": 1140,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.77882635423744e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..083c07b08b24de11301a05b4a540021abebe4c73
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/trainer_state.json
@@ -0,0 +1,1252 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.796139927623643,
+  "eval_steps": 20,
+  "global_step": 1160,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.88810881347584e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..cfc5c24c5735eeb53a154eac71dd475717b69d34
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/trainer_state.json
@@ -0,0 +1,1273 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.844390832328106,
+  "eval_steps": 20,
+  "global_step": 1180,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.9850952863744e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7d4f60a6dd556fa082cddad7d74edfb306d678a9
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/trainer_state.json
@@ -0,0 +1,160 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.28950542822677927,
+  "eval_steps": 20,
+  "global_step": 120,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.1751428171776e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c237abd0bc344e12669fa125199c7e4ce22e420c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/trainer_state.json
@@ -0,0 +1,1294 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.8926417370325694,
+  "eval_steps": 20,
+  "global_step": 1200,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.08848278441984e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..2f8eecfeb16f8a1e073d156f20ca1ee1f3b22350
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/trainer_state.json
@@ -0,0 +1,1315 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.9408926417370327,
+  "eval_steps": 20,
+  "global_step": 1220,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.18975712352256e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e9310127996ee1a84d7686d7f20a2e2da19c72ba
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/trainer_state.json
@@ -0,0 +1,1336 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.989143546441496,
+  "eval_steps": 20,
+  "global_step": 1240,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.28445263122432e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..319b272fa03c0103c459be2641427d3083cb0926
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/trainer_state.json
@@ -0,0 +1,1357 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0361881785283473,
+  "eval_steps": 20,
+  "global_step": 1260,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.38081421772032e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..9424d378812f13d9e27b0d5d80b83aea60dc374c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/trainer_state.json
@@ -0,0 +1,1378 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0844390832328106,
+  "eval_steps": 20,
+  "global_step": 1280,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.49323564121344e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c4380adbfcca66e6d392099fc23de0a4acb2e516
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/trainer_state.json
@@ -0,0 +1,1399 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.132689987937274,
+  "eval_steps": 20,
+  "global_step": 1300,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.60286687425792e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..fed9c0d29347132927d580d472b4a9b3066109d5
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/trainer_state.json
@@ -0,0 +1,1420 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.180940892641737,
+  "eval_steps": 20,
+  "global_step": 1320,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.7166013285504e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e8e6f8dbf579311a4d67df4b173b8155487719a9
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/trainer_state.json
@@ -0,0 +1,1441 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.2291917973462003,
+  "eval_steps": 20,
+  "global_step": 1340,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.81639850800384e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..799a5171f25176bef3ab5a90ff5be9f0d159e151
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/trainer_state.json
@@ -0,0 +1,1462 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.2774427020506636,
+  "eval_steps": 20,
+  "global_step": 1360,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.91881491035392e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..37a9d83eda8dc129c944390b7dcc82ebda547e31
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/trainer_state.json
@@ -0,0 +1,1483 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.325693606755127,
+  "eval_steps": 20,
+  "global_step": 1380,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.01778460685568e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..799f08361355ff18786091b7c81400400e5bb9bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/trainer_state.json
@@ -0,0 +1,181 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.33775633293124246,
+  "eval_steps": 20,
+  "global_step": 140,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.325617668096e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..33c1047b006d3d8ffd4d85e0eddaacf0176207f8
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/trainer_state.json
@@ -0,0 +1,1504 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.37394451145959,
+  "eval_steps": 20,
+  "global_step": 1400,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.11747236707584e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a8ba97c0e113e5a762fd7444330ea62eda819692
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/trainer_state.json
@@ -0,0 +1,1525 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.422195416164053,
+  "eval_steps": 20,
+  "global_step": 1420,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.21436309814528e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..93d879702b473e0599c8e0c8b6a135eefb78050e
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/trainer_state.json
@@ -0,0 +1,1546 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.470446320868516,
+  "eval_steps": 20,
+  "global_step": 1440,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.316492275008e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d80d0ca8009df759f1f86d208d686cde629d8c9d
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/trainer_state.json
@@ -0,0 +1,1567 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.5186972255729794,
+  "eval_steps": 20,
+  "global_step": 1460,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.42217073825024e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e83c981c181e5cb294d9c966e9e203787dbd77ee
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/trainer_state.json
@@ -0,0 +1,1588 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.5669481302774426,
+  "eval_steps": 20,
+  "global_step": 1480,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.52977087677696e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..3e0e06b2bc5eecf5c40edc1e2c402419bf28194a
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/trainer_state.json
@@ -0,0 +1,1609 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.615199034981906,
+  "eval_steps": 20,
+  "global_step": 1500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.62582728619264e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..070444a206cda1e129f93650dd9c695bf895a389
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/trainer_state.json
@@ -0,0 +1,1630 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.663449939686369,
+  "eval_steps": 20,
+  "global_step": 1520,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.72663659355392e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7a97e2df65ef5a95faa39d87e7d152a48084c8ab
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/trainer_state.json
@@ -0,0 +1,1651 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.7117008443908324,
+  "eval_steps": 20,
+  "global_step": 1540,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.833669119808e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e29d11be527d2eaecd4f9dfa639c53545efd4881
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/trainer_state.json
@@ -0,0 +1,1672 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.7599517490952956,
+  "eval_steps": 20,
+  "global_step": 1560,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.92949301335296e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..474247b7b01dda41b59c453ca6f4160fc31ca422
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/trainer_state.json
@@ -0,0 +1,1693 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.808202653799759,
+  "eval_steps": 20,
+  "global_step": 1580,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.02561097108736e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d22618a72b22e1d2e15034531f8b27bbb97d47e2
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/trainer_state.json
@@ -0,0 +1,202 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.38600723763570566,
+  "eval_steps": 20,
+  "global_step": 160,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.311416572928e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c719e62c29ada493d8ec14417f7048f5866c2e6f
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/trainer_state.json
@@ -0,0 +1,1714 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.856453558504222,
+  "eval_steps": 20,
+  "global_step": 1600,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.12746659986688e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..9d429b67a22904f3e636e9f87c91cc48aa2dc8c2
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/trainer_state.json
@@ -0,0 +1,1735 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.9047044632086854,
+  "eval_steps": 20,
+  "global_step": 1620,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.23430764246272e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ed58e37a8ef05fe401087f7be03adef4004877cb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/trainer_state.json
@@ -0,0 +1,1756 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.952955367913148,
+  "eval_steps": 20,
+  "global_step": 1640,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.33664881908992e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..54f8d6c375e4a1318bc114794c3378a26cc8dd0e
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/trainer_state.json
@@ -0,0 +1,1777 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.0,
+  "eval_steps": 20,
+  "global_step": 1660,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.43045871487232e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..2fc38851c7a7f45d761022df87cdaefdc0491b2a
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/trainer_state.json
@@ -0,0 +1,1798 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.048250904704463,
+  "eval_steps": 20,
+  "global_step": 1680,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.53397614825728e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..3cbe2ec2dd92f345942ae3fd4f463dd453e2be93
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/trainer_state.json
@@ -0,0 +1,1819 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.0965018094089265,
+  "eval_steps": 20,
+  "global_step": 1700,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.62847333359872e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c3c7a8717b63e4953647e67cc7c2602e17fcaaed
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/trainer_state.json
@@ -0,0 +1,1840 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.144752714113389,
+  "eval_steps": 20,
+  "global_step": 1720,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.73270883070208e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..12044ed3ca0f976012431d9fc8388d3369ba6469
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/trainer_state.json
@@ -0,0 +1,1861 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.193003618817853,
+  "eval_steps": 20,
+  "global_step": 1740,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.8321367202432e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..f6d0af9ee360c57ec226c54a74051d2986c7f802
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/trainer_state.json
@@ -0,0 +1,1882 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.241254523522316,
+  "eval_steps": 20,
+  "global_step": 1760,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.93257673769216e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..8ed3b3e47d1f83eb2d54a281acc43a40e8377707
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/trainer_state.json
@@ -0,0 +1,1903 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.2895054282267795,
+  "eval_steps": 20,
+  "global_step": 1780,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.03289365850368e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..362b22e585f341d434f9f931c0f5708201f6a500
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/trainer_state.json
@@ -0,0 +1,223 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.43425814234016885,
+  "eval_steps": 20,
+  "global_step": 180,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.3344180170752e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..929f12d02731d0c6a65d44aa1e07e154884308dd
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/trainer_state.json
@@ -0,0 +1,1924 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.337756332931242,
+  "eval_steps": 20,
+  "global_step": 1800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.13988515254528e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..85883a37bf4a6848a3a0abd1b5b87eb11067bd87
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/trainer_state.json
@@ -0,0 +1,1945 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.386007237635706,
+  "eval_steps": 20,
+  "global_step": 1820,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.23728878627072e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a0c879278454eaf5f303e23286f992db8f6ccc1c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/trainer_state.json
@@ -0,0 +1,1966 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.434258142340169,
+  "eval_steps": 20,
+  "global_step": 1840,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.3384058352256e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7540773fb48e0987920a250cd18090a7d6b6033f
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/trainer_state.json
@@ -0,0 +1,1987 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.4825090470446325,
+  "eval_steps": 20,
+  "global_step": 1860,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.4410547534464e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..cb9836a69d93145e2620649be098d6a3fab10fc3
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/trainer_state.json
@@ -0,0 +1,2008 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.530759951749095,
+  "eval_steps": 20,
+  "global_step": 1880,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.542917220928e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..877cdf4eec742e38ba98d504b2291ec1ce3d590c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/trainer_state.json
@@ -0,0 +1,2029 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.579010856453559,
+  "eval_steps": 20,
+  "global_step": 1900,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.63221015398656e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..885d4897f7d514ce3ff1a60095d88ddbfb14c406
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/trainer_state.json
@@ -0,0 +1,2050 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.627261761158022,
+  "eval_steps": 20,
+  "global_step": 1920,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.73311520317696e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d30e007f2fa4467b79ffcdab187a2ba142bd0972
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/trainer_state.json
@@ -0,0 +1,2071 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.675512665862485,
+  "eval_steps": 20,
+  "global_step": 1940,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.83231057684736e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..bff85546e7eb51d2c1eacfde24b52171385b4520
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/trainer_state.json
@@ -0,0 +1,2092 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.723763570566948,
+  "eval_steps": 20,
+  "global_step": 1960,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.93308569069824e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1980/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1980/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1980/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1980/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1980/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1980/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1980/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1980/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1980/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1980/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1980/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1980/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1980/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1980/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..9fd7217e1d1ad2411dccdb4a48a859d4c33cb816
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1980/trainer_state.json
@@ -0,0 +1,2113 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.772014475271411,
+  "eval_steps": 20,
+  "global_step": 1980,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.00234386225792e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-20/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-20/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-20/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-20/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-20/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-20/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-20/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-20/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-20/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-20/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-20/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-20/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-20/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-20/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..17bdf7e6e232e2860e51d8dd9f564af33acec9f1
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-20/trainer_state.json
@@ -0,0 +1,55 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.04825090470446321,
+  "eval_steps": 20,
+  "global_step": 20,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.0192401580032e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-200/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-200/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-200/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-200/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-200/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-200/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-200/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-200/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-200/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-200/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-200/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-200/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-200/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-200/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..5ba3235d45350594c291048e7ca73b2ead2e0f36
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-200/trainer_state.json
@@ -0,0 +1,244 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.4825090470446321,
+  "eval_steps": 20,
+  "global_step": 200,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.03407330281472e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2000/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2000/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2000/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2000/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2000/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2000/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2000/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2000/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2000/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2000/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2000/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2000/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2000/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2000/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e0f3588791892825f2c827d3918df8be86f4125f
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2000/trainer_state.json
@@ -0,0 +1,2134 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.820265379975875,
+  "eval_steps": 20,
+  "global_step": 2000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.012953424664832e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2020/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2020/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2020/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2020/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2020/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2020/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2020/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2020/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2020/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2020/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2020/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2020/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2020/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2020/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..3c8a1e0c5deff7800558a4eb354ecf729715c0be
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2020/trainer_state.json
@@ -0,0 +1,2155 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.868516284680338,
+  "eval_steps": 20,
+  "global_step": 2020,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.023900818954496e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2040/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2040/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2040/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2040/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2040/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2040/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2040/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2040/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2040/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2040/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2040/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2040/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2040/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2040/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..fde45e1cdac1438b1ef2cd436aeb2781b3bfb597
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2040/trainer_state.json
@@ -0,0 +1,2176 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.916767189384801,
+  "eval_steps": 20,
+  "global_step": 2040,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.03357279530624e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2060/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2060/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2060/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2060/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2060/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2060/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2060/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2060/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2060/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2060/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2060/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2060/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2060/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2060/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d70825dc1eb6e5e3e4716274f6683fad95523109
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2060/trainer_state.json
@@ -0,0 +1,2197 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.965018094089264,
+  "eval_steps": 20,
+  "global_step": 2060,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.042724346429696e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2080/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2080/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2080/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2080/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2080/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2080/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2080/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2080/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2080/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2080/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2080/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2080/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2080/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2080/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ddb34155cbd7b669c83fabaa7675bc5a711d52fd
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2080/trainer_state.json
@@ -0,0 +1,2218 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.012062726176116,
+  "eval_steps": 20,
+  "global_step": 2080,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.053239278296576e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2100/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2100/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2100/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2100/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2100/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2100/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2100/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2100/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2100/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2100/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2100/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2100/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2100/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2100/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..4809f7765dda1b9a3c74b295463308339bea2815
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2100/trainer_state.json
@@ -0,0 +1,2239 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.060313630880579,
+  "eval_steps": 20,
+  "global_step": 2100,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.06263428721408e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2120/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2120/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2120/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2120/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2120/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2120/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2120/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2120/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2120/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2120/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2120/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2120/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2120/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2120/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..628c3d39673674570c3b06cd46d418d8e09670c0
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2120/trainer_state.json
@@ -0,0 +1,2260 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.108564535585042,
+  "eval_steps": 20,
+  "global_step": 2120,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.072157863730688e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2140/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2140/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2140/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2140/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2140/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2140/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2140/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2140/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2140/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2140/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2140/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2140/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2140/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2140/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..74e9d35d8f18c94bae43cb9e5820e9f60ac305b1
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2140/trainer_state.json
@@ -0,0 +1,2281 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.156815440289505,
+  "eval_steps": 20,
+  "global_step": 2140,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.081538511373824e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2160/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2160/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2160/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2160/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2160/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2160/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2160/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2160/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2160/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2160/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2160/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2160/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2160/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2160/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c492e9fdedb8632818bf4fa153dead367e87565f
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2160/trainer_state.json
@@ -0,0 +1,2302 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.205066344993969,
+  "eval_steps": 20,
+  "global_step": 2160,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.09116877164288e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2180/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2180/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2180/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2180/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2180/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2180/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2180/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2180/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2180/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2180/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2180/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2180/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2180/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2180/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c80ab8e59f1f1bc252841512e1189411e34deccb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2180/trainer_state.json
@@ -0,0 +1,2323 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.253317249698432,
+  "eval_steps": 20,
+  "global_step": 2180,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.101687379312128e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-220/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-220/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-220/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-220/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-220/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-220/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-220/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-220/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-220/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-220/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-220/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-220/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-220/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-220/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a9a03e37f14d3732a62c05d5ef826363c53fd6d0
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-220/trainer_state.json
@@ -0,0 +1,265 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.5307599517490953,
+  "eval_steps": 20,
+  "global_step": 220,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.12699758667776e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2200/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2200/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2200/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2200/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2200/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2200/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2200/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2200/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2200/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2200/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2200/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2200/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2200/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2200/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..329a25de88eff842aae11101d9279e50bba3ce04
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2200/trainer_state.json
@@ -0,0 +1,2344 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.301568154402895,
+  "eval_steps": 20,
+  "global_step": 2200,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.111436632997376e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2220/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2220/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2220/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2220/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2220/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2220/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2220/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2220/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2220/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2220/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2220/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2220/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2220/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2220/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..eeb4ccb4b8b431cb8b3822eff7874dfadf781203
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2220/trainer_state.json
@@ -0,0 +1,2365 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.349819059107358,
+  "eval_steps": 20,
+  "global_step": 2220,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.12135822197504e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2240/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2240/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2240/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2240/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2240/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2240/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2240/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2240/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2240/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2240/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2240/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2240/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2240/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2240/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c634c71596f568561d5a03df50c96272451be486
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2240/trainer_state.json
@@ -0,0 +1,2386 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.398069963811821,
+  "eval_steps": 20,
+  "global_step": 2240,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.131719539496448e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2260/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2260/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2260/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2260/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2260/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2260/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2260/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2260/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2260/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2260/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2260/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2260/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2260/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2260/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..dbcca27ed3c45ebd957b4510a871762145d35e51
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2260/trainer_state.json
@@ -0,0 +1,2407 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.446320868516285,
+  "eval_steps": 20,
+  "global_step": 2260,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.141223967647232e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2280/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2280/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2280/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2280/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2280/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2280/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2280/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2280/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2280/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2280/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2280/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2280/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2280/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2280/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d3bb88f2daea9a86156fe2849f032fe7f6b17c87
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2280/trainer_state.json
@@ -0,0 +1,2428 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.4945717732207475,
+  "eval_steps": 20,
+  "global_step": 2280,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.150872692411904e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2300/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2300/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2300/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2300/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2300/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2300/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2300/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2300/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2300/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2300/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2300/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2300/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2300/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2300/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..108664464b2fafbbb80eb0e441264444a3fe3635
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2300/trainer_state.json
@@ -0,0 +1,2449 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.542822677925211,
+  "eval_steps": 20,
+  "global_step": 2300,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.161259997001216e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2320/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2320/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2320/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2320/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2320/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2320/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2320/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2320/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2320/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2320/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2320/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2320/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2320/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2320/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e1449379270b7e6d85f81008ed75de888a7c89b9
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2320/trainer_state.json
@@ -0,0 +1,2470 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.591073582629674,
+  "eval_steps": 20,
+  "global_step": 2320,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.172979480755712e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2340/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2340/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2340/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2340/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2340/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2340/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2340/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2340/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2340/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2340/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2340/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2340/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2340/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2340/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a295a06edc9a31f0c0e1d33a7a8d046eb2efb2d4
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2340/trainer_state.json
@@ -0,0 +1,2491 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.639324487334138,
+  "eval_steps": 20,
+  "global_step": 2340,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.184396010008064e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2360/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2360/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2360/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2360/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2360/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2360/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2360/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2360/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2360/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2360/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2360/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2360/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2360/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2360/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..4dad6d3106e81616e0e8f3ae46a1880ad304f4f1
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2360/trainer_state.json
@@ -0,0 +1,2512 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.6875753920386005,
+  "eval_steps": 20,
+  "global_step": 2360,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.194835972603392e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2380/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2380/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2380/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2380/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2380/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2380/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2380/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2380/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2380/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2380/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2380/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2380/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2380/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2380/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b24676fd7b207fabebe9f3d0e0884e4b981094bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2380/trainer_state.json
@@ -0,0 +1,2533 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.735826296743064,
+  "eval_steps": 20,
+  "global_step": 2380,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.20520070947584e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-240/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-240/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-240/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-240/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-240/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-240/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-240/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-240/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-240/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-240/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-240/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-240/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-240/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-240/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..4d89c47f91e068efdf705485ca558c0fcd4912b9
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-240/trainer_state.json
@@ -0,0 +1,286 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.5790108564535585,
+  "eval_steps": 20,
+  "global_step": 240,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.22550225143808e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2400/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2400/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2400/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2400/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2400/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2400/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2400/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2400/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2400/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2400/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2400/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2400/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2400/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2400/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..5aac8b7e0627badc3ce6334f1fda834d56cb987d
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2400/trainer_state.json
@@ -0,0 +1,2554 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.784077201447527,
+  "eval_steps": 20,
+  "global_step": 2400,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.215871136331264e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2420/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2420/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2420/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2420/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2420/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2420/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2420/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2420/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2420/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2420/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2420/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2420/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2420/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2420/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..95e1d25d44535f702a815faea5bb4360ac27edc6
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2420/trainer_state.json
@@ -0,0 +1,2575 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.832328106151991,
+  "eval_steps": 20,
+  "global_step": 2420,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.225269564599808e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2440/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2440/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2440/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2440/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2440/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2440/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2440/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2440/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2440/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2440/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2440/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2440/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2440/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2440/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..dee2e7b81a80de354a3d15386ba803ddf5ff899a
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2440/trainer_state.json
@@ -0,0 +1,2596 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.8805790108564535,
+  "eval_steps": 20,
+  "global_step": 2440,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.235377166274048e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2460/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2460/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2460/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2460/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2460/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2460/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2460/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2460/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2460/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2460/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2460/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2460/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2460/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2460/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae4f073ec371f4d47682460e9b25a6703f71e4b6
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2460/trainer_state.json
@@ -0,0 +1,2617 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.928829915560916,
+  "eval_steps": 20,
+  "global_step": 2460,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.245342522945024e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2480/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2480/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2480/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2480/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2480/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2480/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2480/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2480/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2480/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2480/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2480/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2480/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2480/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2480/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..affc5114a5d7c6f3c66af72d6702ed64613d1c28
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2480/trainer_state.json
@@ -0,0 +1,2638 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.97708082026538,
+  "eval_steps": 20,
+  "global_step": 2480,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.255260692571648e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2500/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2500/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2500/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2500/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2500/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2500/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2500/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2500/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2500/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2500/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2500/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2500/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2500/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2500/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c3ae3f1bb6ec78f3688e206e00c65e74ca0f33d
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2500/trainer_state.json
@@ -0,0 +1,2659 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.024125452352232,
+  "eval_steps": 20,
+  "global_step": 2500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.266128245014528e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2520/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2520/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2520/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2520/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2520/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2520/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2520/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2520/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2520/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2520/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2520/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2520/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2520/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2520/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..49e893d3a48cc056da9b8d549b7cf57b3c943e33
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2520/trainer_state.json
@@ -0,0 +1,2680 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.072376357056695,
+  "eval_steps": 20,
+  "global_step": 2520,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.27587271160832e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2540/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2540/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2540/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2540/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2540/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2540/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2540/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2540/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2540/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2540/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2540/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2540/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2540/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2540/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e4e9a635a6b551908177e813843db46463238c6
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2540/trainer_state.json
@@ -0,0 +1,2701 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.120627261761158,
+  "eval_steps": 20,
+  "global_step": 2540,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.285468094496768e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2560/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2560/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2560/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2560/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2560/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2560/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2560/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2560/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2560/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2560/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2560/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2560/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2560/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2560/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..cc28f40a3dd78700466bbcb3e94eba8440195001
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2560/trainer_state.json
@@ -0,0 +1,2722 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.168878166465621,
+  "eval_steps": 20,
+  "global_step": 2560,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.294447994198016e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2580/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2580/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2580/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2580/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2580/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2580/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2580/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2580/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2580/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2580/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2580/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2580/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2580/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2580/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..acf25d1d142a73d7631e3b97fc5c59dc8e60839c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2580/trainer_state.json
@@ -0,0 +1,2743 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.217129071170085,
+  "eval_steps": 20,
+  "global_step": 2580,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.30466022801408e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-260/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-260/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-260/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-260/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-260/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-260/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-260/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-260/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-260/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-260/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-260/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-260/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-260/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-260/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..21650d4109e50556393b33b86e644342a6b30857
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-260/trainer_state.json
@@ -0,0 +1,307 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.6272617611580217,
+  "eval_steps": 20,
+  "global_step": 260,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.32497117319168e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2600/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2600/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2600/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2600/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2600/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2600/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2600/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2600/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2600/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2600/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2600/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2600/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2600/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2600/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc40ea1969082f31940f29f0bec7512843f5763b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2600/trainer_state.json
@@ -0,0 +1,2764 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.265379975874548,
+  "eval_steps": 20,
+  "global_step": 2600,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.316223789361152e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2620/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2620/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2620/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2620/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2620/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2620/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2620/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2620/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2620/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2620/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2620/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2620/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2620/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2620/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b97e279231d1d2788cc4c20c3da9acd0ab8c1a14
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2620/trainer_state.json
@@ -0,0 +1,2785 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.3136308805790105,
+  "eval_steps": 20,
+  "global_step": 2620,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.327587660607488e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2640/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2640/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2640/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2640/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2640/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2640/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2640/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2640/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2640/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2640/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2640/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2640/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2640/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2640/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..15806a9127dbc094e5e89cb67977db95868693be
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2640/trainer_state.json
@@ -0,0 +1,2806 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.361881785283474,
+  "eval_steps": 20,
+  "global_step": 2640,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.338233468135424e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2660/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2660/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2660/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2660/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2660/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2660/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2660/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2660/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2660/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2660/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2660/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2660/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2660/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2660/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c51b1c37cd97e5e7f8abc7b49490eb4b420f7c6
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2660/trainer_state.json
@@ -0,0 +1,2827 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.410132689987937,
+  "eval_steps": 20,
+  "global_step": 2660,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.347335096733696e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2680/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2680/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2680/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2680/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2680/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2680/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2680/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2680/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2680/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2680/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2680/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2680/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2680/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2680/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..04051244b5e9fb150fb7497a37315fa5fe0a0a37
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2680/trainer_state.json
@@ -0,0 +1,2848 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.458383594692401,
+  "eval_steps": 20,
+  "global_step": 2680,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.357896788226048e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2700/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2700/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2700/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2700/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2700/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2700/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2700/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2700/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2700/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2700/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2700/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2700/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2700/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2700/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b599b0a0b685dda72bd5ede38e30644afb01fd0e
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2700/trainer_state.json
@@ -0,0 +1,2869 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.5066344993968634,
+  "eval_steps": 20,
+  "global_step": 2700,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.367404635727872e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2720/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2720/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2720/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2720/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2720/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2720/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2720/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2720/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2720/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2720/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2720/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2720/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2720/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2720/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..351c06e6c9dcedd5d2ebbd87ebbc4df022ff4146
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2720/trainer_state.json
@@ -0,0 +1,2890 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.554885404101327,
+  "eval_steps": 20,
+  "global_step": 2720,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.37645634180096e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2740/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2740/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2740/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2740/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2740/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2740/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2740/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2740/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2740/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2740/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2740/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2740/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2740/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2740/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..bc7cf8ddbd75dfe07fbe1d13c086cb8d0932d23c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2740/trainer_state.json
@@ -0,0 +1,2911 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.60313630880579,
+  "eval_steps": 20,
+  "global_step": 2740,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.386491453233152e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2760/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2760/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2760/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2760/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2760/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2760/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2760/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2760/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2760/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2760/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2760/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2760/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2760/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2760/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..15516abe168149fd3218e2e162af3f0a8a5b6888
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2760/trainer_state.json
@@ -0,0 +1,2932 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.651387213510254,
+  "eval_steps": 20,
+  "global_step": 2760,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.396458861514752e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2780/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2780/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2780/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2780/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2780/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2780/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2780/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2780/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2780/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2780/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2780/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2780/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2780/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2780/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..56f19bd81d83fa8a749a2f2c505caf1b27dbc26a
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2780/trainer_state.json
@@ -0,0 +1,2953 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.699638118214716,
+  "eval_steps": 20,
+  "global_step": 2780,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.40622179260416e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-280/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-280/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-280/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-280/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-280/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-280/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-280/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-280/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-280/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-280/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-280/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-280/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-280/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-280/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..3b7938b0a77609cf711579fdf6a15d8f2765353a
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-280/trainer_state.json
@@ -0,0 +1,328 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.6755126658624849,
+  "eval_steps": 20,
+  "global_step": 280,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.43159337732096e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2800/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2800/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2800/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2800/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2800/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2800/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2800/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2800/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2800/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2800/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2800/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2800/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2800/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2800/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..9563b5a1eda5890180b11da7024c351b7ba8a23e
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2800/trainer_state.json
@@ -0,0 +1,2974 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.74788902291918,
+  "eval_steps": 20,
+  "global_step": 2800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.416029175257088e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2820/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2820/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2820/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2820/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2820/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2820/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2820/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2820/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2820/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2820/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2820/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2820/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2820/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2820/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..1260c8ab7a4dd8f526c195c5b3d51afc7419176b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2820/trainer_state.json
@@ -0,0 +1,2995 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.796139927623643,
+  "eval_steps": 20,
+  "global_step": 2820,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.426538892613632e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2840/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2840/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2840/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2840/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2840/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2840/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2840/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2840/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2840/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2840/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2840/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2840/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2840/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2840/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b634516db5af61ba15cf5d084e21af59a442eae5
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2840/trainer_state.json
@@ -0,0 +1,3016 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.844390832328106,
+  "eval_steps": 20,
+  "global_step": 2840,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.436608197556224e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2860/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2860/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2860/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2860/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2860/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2860/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2860/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2860/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2860/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2860/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2860/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2860/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2860/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2860/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7c9327e88467d020f5f61179aee5687e0d9ee87f
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2860/trainer_state.json
@@ -0,0 +1,3037 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.892641737032569,
+  "eval_steps": 20,
+  "global_step": 2860,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.446390277011456e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2880/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2880/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2880/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2880/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2880/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2880/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2880/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2880/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2880/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2880/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2880/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2880/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2880/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2880/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d92dd280fd54dad1b43fad8ad7e0c6604b626bbc
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2880/trainer_state.json
@@ -0,0 +1,3058 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.940892641737032,
+  "eval_steps": 20,
+  "global_step": 2880,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.456912987901952e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2900/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2900/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2900/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2900/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2900/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2900/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2900/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2900/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2900/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2900/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2900/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2900/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2900/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2900/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..0ad1de230808d0ab6671e7660e3eeb312015c8c1
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2900/trainer_state.json
@@ -0,0 +1,3079 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.989143546441496,
+  "eval_steps": 20,
+  "global_step": 2900,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.46702811214848e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2920/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2920/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2920/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2920/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2920/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2920/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2920/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2920/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2920/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2920/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2920/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2920/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2920/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2920/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..2d8a854e78daf57cfea1f1a40209e8c8882f3311
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2920/trainer_state.json
@@ -0,0 +1,3100 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.036188178528348,
+  "eval_steps": 20,
+  "global_step": 2920,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.477191363760896e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2940/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2940/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2940/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2940/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2940/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2940/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2940/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2940/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2940/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2940/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2940/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2940/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2940/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2940/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..525c0baa271832ce06ac557090d22a61deac7b5b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2940/trainer_state.json
@@ -0,0 +1,3121 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.084439083232811,
+  "eval_steps": 20,
+  "global_step": 2940,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.486946088407808e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2960/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2960/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2960/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2960/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2960/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2960/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2960/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2960/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2960/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2960/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2960/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2960/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2960/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2960/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d10449fb5a7a46de4eaffe9fddb33b75c2b743cc
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2960/trainer_state.json
@@ -0,0 +1,3142 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.132689987937274,
+  "eval_steps": 20,
+  "global_step": 2960,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.49675210332032e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2980/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2980/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2980/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2980/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2980/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2980/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2980/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2980/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2980/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2980/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2980/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2980/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2980/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2980/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d05e76ae9101b6e40600773b61a0f647308f4762
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-2980/trainer_state.json
@@ -0,0 +1,3163 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.180940892641737,
+  "eval_steps": 20,
+  "global_step": 2980,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.5068460275904e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-300/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-300/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-300/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-300/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-300/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-300/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-300/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-300/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-300/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-300/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-300/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-300/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-300/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-300/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..530c34aff7dc72534a043830f6da4ee62d4049d1
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-300/trainer_state.json
@@ -0,0 +1,349 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.7237635705669482,
+  "eval_steps": 20,
+  "global_step": 300,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.53407816669184e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3000/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3000/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3000/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3000/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3000/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3000/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3000/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3000/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3000/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3000/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3000/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3000/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3000/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3000/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..539be4cc8e25db0d427fab1e62bb4f3897c81895
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3000/trainer_state.json
@@ -0,0 +1,3184 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.2291917973462,
+  "eval_steps": 20,
+  "global_step": 3000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.517889847579392e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3020/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3020/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3020/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3020/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3020/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3020/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3020/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3020/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3020/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3020/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3020/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3020/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3020/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3020/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..2a444ae2fa5c1eaa781ea65052131d85b059057f
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3020/trainer_state.json
@@ -0,0 +1,3205 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.277442702050664,
+  "eval_steps": 20,
+  "global_step": 3020,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.528316816640768e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3040/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3040/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3040/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3040/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3040/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3040/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3040/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3040/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3040/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3040/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3040/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3040/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3040/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3040/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a63c381e141e314f5a4898739dff43ddb086397c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3040/trainer_state.json
@@ -0,0 +1,3226 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.325693606755126,
+  "eval_steps": 20,
+  "global_step": 3040,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.538312947471104e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3060/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3060/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3060/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3060/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3060/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3060/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3060/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3060/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3060/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3060/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3060/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3060/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3060/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3060/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..96e70a8b5dde3e5938ab44e37ed8995f372a5e75
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3060/trainer_state.json
@@ -0,0 +1,3247 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.37394451145959,
+  "eval_steps": 20,
+  "global_step": 3060,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.548778213264128e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3080/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3080/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3080/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3080/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3080/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3080/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3080/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3080/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3080/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3080/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3080/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3080/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3080/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3080/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c2b295fc04aa83d722462ea4a8cc33e220fe4f6
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3080/trainer_state.json
@@ -0,0 +1,3268 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.422195416164053,
+  "eval_steps": 20,
+  "global_step": 3080,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.55879622794112e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3100/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3100/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3100/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3100/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3100/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3100/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3100/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3100/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3100/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3100/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3100/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3100/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3100/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3100/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c09ada2a9d6d368fbbe9bce91a69559bc13c88ab
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3100/trainer_state.json
@@ -0,0 +1,3289 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.470446320868517,
+  "eval_steps": 20,
+  "global_step": 3100,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.56810028212096e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3120/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3120/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3120/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3120/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3120/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3120/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3120/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3120/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3120/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3120/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3120/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3120/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3120/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3120/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..773b43ea513de6b22ad6f12c4658834219060973
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3120/trainer_state.json
@@ -0,0 +1,3310 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.518697225572979,
+  "eval_steps": 20,
+  "global_step": 3120,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.578918424941312e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3140/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3140/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3140/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3140/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3140/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3140/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3140/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3140/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3140/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3140/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3140/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3140/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3140/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3140/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..3d5f96b9ea45c7322d949c02ed7ad3c1b0bffc8e
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3140/trainer_state.json
@@ -0,0 +1,3331 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.566948130277443,
+  "eval_steps": 20,
+  "global_step": 3140,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.589942412694272e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3160/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3160/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3160/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3160/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3160/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3160/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3160/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3160/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3160/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3160/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3160/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3160/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3160/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3160/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..859bd1ae08199cc707dbe824249f2c5499d00ec8
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3160/trainer_state.json
@@ -0,0 +1,3352 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.615199034981906,
+  "eval_steps": 20,
+  "global_step": 3160,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.60040973009792e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3180/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3180/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3180/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3180/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3180/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3180/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3180/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3180/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3180/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3180/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3180/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3180/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3180/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3180/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..3302d4c78cd3536c369ce5fa2b64e7beaf169fc0
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3180/trainer_state.json
@@ -0,0 +1,3373 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.66344993968637,
+  "eval_steps": 20,
+  "global_step": 3180,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.610287551382272e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-320/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-320/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-320/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-320/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-320/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-320/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-320/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-320/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-320/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-320/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-320/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-320/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-320/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-320/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..84646dbb0699c7f765461b8117092fa9b65cd660
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-320/trainer_state.json
@@ -0,0 +1,370 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.7720144752714113,
+  "eval_steps": 20,
+  "global_step": 320,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.64476939855872e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3200/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3200/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3200/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3200/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3200/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3200/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3200/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3200/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3200/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3200/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3200/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3200/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3200/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3200/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a420dce5cc32f73855521ad0ce522777c4d616b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3200/trainer_state.json
@@ -0,0 +1,3394 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.711700844390832,
+  "eval_steps": 20,
+  "global_step": 3200,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.619632637774592e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3220/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3220/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3220/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3220/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3220/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3220/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3220/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3220/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3220/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3220/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3220/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3220/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3220/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3220/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ffeaec7e515afb710bd36507f4d0e390edaae7fa
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3220/trainer_state.json
@@ -0,0 +1,3415 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.759951749095295,
+  "eval_steps": 20,
+  "global_step": 3220,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.629993271425792e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3240/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3240/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3240/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3240/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3240/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3240/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3240/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3240/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3240/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3240/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3240/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3240/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3240/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3240/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..86bee22ea05ee9986df6d6da03e426bee085adaf
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3240/trainer_state.json
@@ -0,0 +1,3436 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.808202653799759,
+  "eval_steps": 20,
+  "global_step": 3240,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.640726614340352e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3260/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3260/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3260/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3260/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3260/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3260/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3260/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3260/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3260/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3260/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3260/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3260/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3260/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3260/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..02506763126bb01ae6c661008e2e2c440ebbee35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3260/trainer_state.json
@@ -0,0 +1,3457 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.856453558504222,
+  "eval_steps": 20,
+  "global_step": 3260,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.651297880015616e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3280/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3280/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3280/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3280/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3280/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3280/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3280/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3280/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3280/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3280/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3280/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3280/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3280/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3280/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b7d9fdf22e902941be1d86bf1bfc94305db4f4d4
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3280/trainer_state.json
@@ -0,0 +1,3478 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.904704463208685,
+  "eval_steps": 20,
+  "global_step": 3280,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.66104850144128e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3300/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3300/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3300/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3300/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3300/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3300/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3300/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3300/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3300/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3300/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3300/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3300/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3300/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3300/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..181e0ab1089df969180ea9027f472e0bccae44e3
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3300/trainer_state.json
@@ -0,0 +1,3499 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.952955367913148,
+  "eval_steps": 20,
+  "global_step": 3300,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.671065832248064e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3320/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3320/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3320/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3320/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3320/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3320/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3320/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3320/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3320/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3320/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3320/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3320/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3320/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3320/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..de3405c15700d4532fbbcabc792f5f6fc5bc256e
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3320/trainer_state.json
@@ -0,0 +1,3520 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.0,
+  "eval_steps": 20,
+  "global_step": 3320,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.680218836595712e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3340/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3340/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3340/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3340/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3340/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3340/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3340/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3340/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3340/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3340/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3340/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3340/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3340/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3340/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ea2b8811562722e04bb1354dc5d1ed2d0f204a0b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3340/trainer_state.json
@@ -0,0 +1,3541 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.048250904704464,
+  "eval_steps": 20,
+  "global_step": 3340,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.689662400297984e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3360/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3360/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3360/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3360/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3360/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3360/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3360/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3360/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3360/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3360/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3360/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3360/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3360/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3360/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..2616b2c2c727fb0a98a763aa1ed55ff57d264a89
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3360/trainer_state.json
@@ -0,0 +1,3562 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.096501809408926,
+  "eval_steps": 20,
+  "global_step": 3360,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.69890627389952e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3380/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3380/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3380/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3380/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3380/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3380/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3380/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3380/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3380/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3380/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3380/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3380/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3380/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3380/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e967d58480dacde066c2786b7d556b958ec50e91
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3380/trainer_state.json
@@ -0,0 +1,3583 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.14475271411339,
+  "eval_steps": 20,
+  "global_step": 3380,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.709091836777472e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-340/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-340/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-340/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-340/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-340/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-340/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-340/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-340/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-340/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-340/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-340/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-340/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-340/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-340/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ac0b0dc6280da5dceb27ca394f91de5fd7dbe18a
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-340/trainer_state.json
@@ -0,0 +1,391 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.8202653799758746,
+  "eval_steps": 20,
+  "global_step": 340,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.74929212114944e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3400/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3400/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3400/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3400/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3400/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3400/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3400/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3400/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3400/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3400/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3400/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3400/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3400/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3400/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..8b73554952a8cc57491d61197012efb067cbb335
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3400/trainer_state.json
@@ -0,0 +1,3604 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.193003618817853,
+  "eval_steps": 20,
+  "global_step": 3400,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.72045844350464e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3420/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3420/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3420/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3420/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3420/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3420/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3420/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3420/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3420/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3420/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3420/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3420/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3420/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3420/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..91647b63ede178e454fe1b9874a6b271ca768175
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3420/trainer_state.json
@@ -0,0 +1,3625 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.241254523522317,
+  "eval_steps": 20,
+  "global_step": 3420,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.730541425851392e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c4a3c82b11c68ea5e61620a3266640f1a13de48f
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/trainer_state.json
@@ -0,0 +1,3646 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.289505428226779,
+  "eval_steps": 20,
+  "global_step": 3440,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.739404383747072e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..5ac50e18d0b2d7f56823aa5a59658df0fee9b88d
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/trainer_state.json
@@ -0,0 +1,3667 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.337756332931242,
+  "eval_steps": 20,
+  "global_step": 3460,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.74928357277184e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b6231e626e23f0a8e5d4566e5b27b96b9de18b72
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/trainer_state.json
@@ -0,0 +1,3688 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.386007237635706,
+  "eval_steps": 20,
+  "global_step": 3480,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.760041535013888e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..65468522342bae1692305592e6af1fe2ead3218c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/trainer_state.json
@@ -0,0 +1,3709 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.43425814234017,
+  "eval_steps": 20,
+  "global_step": 3500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.770493123402752e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c3d5209feef752e0af0b62d6c984413256115af1
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/trainer_state.json
@@ -0,0 +1,3730 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.482509047044632,
+  "eval_steps": 20,
+  "global_step": 3520,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.78057063478784e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..0e32462509ecc670c90d9510f4313228bfd095cb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/trainer_state.json
@@ -0,0 +1,3751 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.530759951749095,
+  "eval_steps": 20,
+  "global_step": 3540,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.79102701026816e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a7802ba196907b75962ccb9defcb275fdc5cb949
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/trainer_state.json
@@ -0,0 +1,3772 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.579010856453559,
+  "eval_steps": 20,
+  "global_step": 3560,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.8009636443904e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b478cc1bbb076b86eb43ba7ac6561b600de54fe4
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/trainer_state.json
@@ -0,0 +1,3793 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.627261761158021,
+  "eval_steps": 20,
+  "global_step": 3580,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.811924716084224e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c296aafbb1054449f1b900dab04a62587b705ced
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/trainer_state.json
@@ -0,0 +1,412 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.8685162846803377,
+  "eval_steps": 20,
+  "global_step": 360,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.84696930295808e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..5548652de919a03d8f7b89ffa14f85351e45a999
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/trainer_state.json
@@ -0,0 +1,3814 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.675512665862485,
+  "eval_steps": 20,
+  "global_step": 3600,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.822174562761728e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..1a566fe2cc52545b2fadefc01e6e8722505fe1c9
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/trainer_state.json
@@ -0,0 +1,3835 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.723763570566948,
+  "eval_steps": 20,
+  "global_step": 3620,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.83264051242496e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..fd59b9feaa4dbfd1858346aeaa6ec1baf6ef365d
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/trainer_state.json
@@ -0,0 +1,3856 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.772014475271412,
+  "eval_steps": 20,
+  "global_step": 3640,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.843316410242048e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a9f326e6049e7d8cd0faca05a7d2dce649c79c64
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/trainer_state.json
@@ -0,0 +1,3877 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.820265379975874,
+  "eval_steps": 20,
+  "global_step": 3660,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.852651922451456e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d79b5fc24475e1e26c26bb46024e6cb8cfdd1f87
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/trainer_state.json
@@ -0,0 +1,3898 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.868516284680338,
+  "eval_steps": 20,
+  "global_step": 3680,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.863077523772416e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..142c2eeda026acebd72320823a8f5a0e42f49730
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/trainer_state.json
@@ -0,0 +1,3919 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.916767189384801,
+  "eval_steps": 20,
+  "global_step": 3700,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.872863022578688e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ae5630f2375b7aa5492b100b50b5387b0348554
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/trainer_state.json
@@ -0,0 +1,3940 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.965018094089265,
+  "eval_steps": 20,
+  "global_step": 3720,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.883224340100096e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..61a67db77cd28009bbae37243dd5d0ee8738a028
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/trainer_state.json
@@ -0,0 +1,3961 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.012062726176115,
+  "eval_steps": 20,
+  "global_step": 3740,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    },
+    {
+      "entropy": 0.17359357613783616,
+      "epoch": 9.012062726176115,
+      "grad_norm": 0.37213996052742004,
+      "learning_rate": 2.138437593000977e-06,
+      "loss": 0.10089871883392335,
+      "mean_token_accuracy": 0.9696638874518566,
+      "num_tokens": 9122597.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.012062726176115,
+      "eval_entropy": 0.2875421148505104,
+      "eval_loss": 0.9357590079307556,
+      "eval_mean_token_accuracy": 0.8264994859025719,
+      "eval_num_tokens": 9122597.0,
+      "eval_runtime": 98.523,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3740
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.892632342551552e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d81a359cd2596cc22afa9fc2a91cf5611cfe3847
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/trainer_state.json
@@ -0,0 +1,3982 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.060313630880579,
+  "eval_steps": 20,
+  "global_step": 3760,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    },
+    {
+      "entropy": 0.17359357613783616,
+      "epoch": 9.012062726176115,
+      "grad_norm": 0.37213996052742004,
+      "learning_rate": 2.138437593000977e-06,
+      "loss": 0.10089871883392335,
+      "mean_token_accuracy": 0.9696638874518566,
+      "num_tokens": 9122597.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.012062726176115,
+      "eval_entropy": 0.2875421148505104,
+      "eval_loss": 0.9357590079307556,
+      "eval_mean_token_accuracy": 0.8264994859025719,
+      "eval_num_tokens": 9122597.0,
+      "eval_runtime": 98.523,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3740
+    },
+    {
+      "entropy": 0.1642639346420765,
+      "epoch": 9.060313630880579,
+      "grad_norm": 0.5511618852615356,
+      "learning_rate": 1.937215494938237e-06,
+      "loss": 0.08524587750434875,
+      "mean_token_accuracy": 0.9754035726189614,
+      "num_tokens": 9171477.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.060313630880579,
+      "eval_entropy": 0.2820046850469675,
+      "eval_loss": 0.9588018655776978,
+      "eval_mean_token_accuracy": 0.8257443647036392,
+      "eval_num_tokens": 9171477.0,
+      "eval_runtime": 98.3992,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3760
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.902099841711104e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..5b0bb8afed415455e1218e95708bcfa4cd9236ab
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/trainer_state.json
@@ -0,0 +1,4003 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.108564535585042,
+  "eval_steps": 20,
+  "global_step": 3780,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    },
+    {
+      "entropy": 0.17359357613783616,
+      "epoch": 9.012062726176115,
+      "grad_norm": 0.37213996052742004,
+      "learning_rate": 2.138437593000977e-06,
+      "loss": 0.10089871883392335,
+      "mean_token_accuracy": 0.9696638874518566,
+      "num_tokens": 9122597.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.012062726176115,
+      "eval_entropy": 0.2875421148505104,
+      "eval_loss": 0.9357590079307556,
+      "eval_mean_token_accuracy": 0.8264994859025719,
+      "eval_num_tokens": 9122597.0,
+      "eval_runtime": 98.523,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3740
+    },
+    {
+      "entropy": 0.1642639346420765,
+      "epoch": 9.060313630880579,
+      "grad_norm": 0.5511618852615356,
+      "learning_rate": 1.937215494938237e-06,
+      "loss": 0.08524587750434875,
+      "mean_token_accuracy": 0.9754035726189614,
+      "num_tokens": 9171477.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.060313630880579,
+      "eval_entropy": 0.2820046850469675,
+      "eval_loss": 0.9588018655776978,
+      "eval_mean_token_accuracy": 0.8257443647036392,
+      "eval_num_tokens": 9171477.0,
+      "eval_runtime": 98.3992,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1630103051662445,
+      "epoch": 9.108564535585042,
+      "grad_norm": 0.5125291347503662,
+      "learning_rate": 1.7456739289470143e-06,
+      "loss": 0.0895907461643219,
+      "mean_token_accuracy": 0.973100857436657,
+      "num_tokens": 9217543.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.108564535585042,
+      "eval_entropy": 0.2786683516388529,
+      "eval_loss": 0.9743365049362183,
+      "eval_mean_token_accuracy": 0.8253507131940863,
+      "eval_num_tokens": 9217543.0,
+      "eval_runtime": 98.3882,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3780
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.911165225188352e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..df15a08bb9bba8458c37c130d66711d3d8a0a2d3
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/trainer_state.json
@@ -0,0 +1,433 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.916767189384801,
+  "eval_steps": 20,
+  "global_step": 380,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.93760262162432e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..12403c3cda0f6533628e09e2fdd9dee37b0bed9c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/trainer_state.json
@@ -0,0 +1,4024 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.156815440289506,
+  "eval_steps": 20,
+  "global_step": 3800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    },
+    {
+      "entropy": 0.17359357613783616,
+      "epoch": 9.012062726176115,
+      "grad_norm": 0.37213996052742004,
+      "learning_rate": 2.138437593000977e-06,
+      "loss": 0.10089871883392335,
+      "mean_token_accuracy": 0.9696638874518566,
+      "num_tokens": 9122597.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.012062726176115,
+      "eval_entropy": 0.2875421148505104,
+      "eval_loss": 0.9357590079307556,
+      "eval_mean_token_accuracy": 0.8264994859025719,
+      "eval_num_tokens": 9122597.0,
+      "eval_runtime": 98.523,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3740
+    },
+    {
+      "entropy": 0.1642639346420765,
+      "epoch": 9.060313630880579,
+      "grad_norm": 0.5511618852615356,
+      "learning_rate": 1.937215494938237e-06,
+      "loss": 0.08524587750434875,
+      "mean_token_accuracy": 0.9754035726189614,
+      "num_tokens": 9171477.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.060313630880579,
+      "eval_entropy": 0.2820046850469675,
+      "eval_loss": 0.9588018655776978,
+      "eval_mean_token_accuracy": 0.8257443647036392,
+      "eval_num_tokens": 9171477.0,
+      "eval_runtime": 98.3992,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1630103051662445,
+      "epoch": 9.108564535585042,
+      "grad_norm": 0.5125291347503662,
+      "learning_rate": 1.7456739289470143e-06,
+      "loss": 0.0895907461643219,
+      "mean_token_accuracy": 0.973100857436657,
+      "num_tokens": 9217543.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.108564535585042,
+      "eval_entropy": 0.2786683516388529,
+      "eval_loss": 0.9743365049362183,
+      "eval_mean_token_accuracy": 0.8253507131940863,
+      "eval_num_tokens": 9217543.0,
+      "eval_runtime": 98.3882,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3780
+    },
+    {
+      "entropy": 0.15873534101992845,
+      "epoch": 9.156815440289506,
+      "grad_norm": 0.5697775483131409,
+      "learning_rate": 1.5638670990247788e-06,
+      "loss": 0.08552997708320617,
+      "mean_token_accuracy": 0.973672679066658,
+      "num_tokens": 9268724.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.156815440289506,
+      "eval_entropy": 0.2781428360704626,
+      "eval_loss": 0.9790021777153015,
+      "eval_mean_token_accuracy": 0.8251898780967413,
+      "eval_num_tokens": 9268724.0,
+      "eval_runtime": 98.5449,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 3800
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.922020296999936e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..869d8132b03e9ec3744304bf6e0c7534433cf998
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/trainer_state.json
@@ -0,0 +1,4045 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.205066344993968,
+  "eval_steps": 20,
+  "global_step": 3820,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    },
+    {
+      "entropy": 0.17359357613783616,
+      "epoch": 9.012062726176115,
+      "grad_norm": 0.37213996052742004,
+      "learning_rate": 2.138437593000977e-06,
+      "loss": 0.10089871883392335,
+      "mean_token_accuracy": 0.9696638874518566,
+      "num_tokens": 9122597.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.012062726176115,
+      "eval_entropy": 0.2875421148505104,
+      "eval_loss": 0.9357590079307556,
+      "eval_mean_token_accuracy": 0.8264994859025719,
+      "eval_num_tokens": 9122597.0,
+      "eval_runtime": 98.523,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3740
+    },
+    {
+      "entropy": 0.1642639346420765,
+      "epoch": 9.060313630880579,
+      "grad_norm": 0.5511618852615356,
+      "learning_rate": 1.937215494938237e-06,
+      "loss": 0.08524587750434875,
+      "mean_token_accuracy": 0.9754035726189614,
+      "num_tokens": 9171477.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.060313630880579,
+      "eval_entropy": 0.2820046850469675,
+      "eval_loss": 0.9588018655776978,
+      "eval_mean_token_accuracy": 0.8257443647036392,
+      "eval_num_tokens": 9171477.0,
+      "eval_runtime": 98.3992,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1630103051662445,
+      "epoch": 9.108564535585042,
+      "grad_norm": 0.5125291347503662,
+      "learning_rate": 1.7456739289470143e-06,
+      "loss": 0.0895907461643219,
+      "mean_token_accuracy": 0.973100857436657,
+      "num_tokens": 9217543.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.108564535585042,
+      "eval_entropy": 0.2786683516388529,
+      "eval_loss": 0.9743365049362183,
+      "eval_mean_token_accuracy": 0.8253507131940863,
+      "eval_num_tokens": 9217543.0,
+      "eval_runtime": 98.3882,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3780
+    },
+    {
+      "entropy": 0.15873534101992845,
+      "epoch": 9.156815440289506,
+      "grad_norm": 0.5697775483131409,
+      "learning_rate": 1.5638670990247788e-06,
+      "loss": 0.08552997708320617,
+      "mean_token_accuracy": 0.973672679066658,
+      "num_tokens": 9268724.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.156815440289506,
+      "eval_entropy": 0.2781428360704626,
+      "eval_loss": 0.9790021777153015,
+      "eval_mean_token_accuracy": 0.8251898780967413,
+      "eval_num_tokens": 9268724.0,
+      "eval_runtime": 98.5449,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 3800
+    },
+    {
+      "entropy": 0.1606229657307267,
+      "epoch": 9.205066344993968,
+      "grad_norm": 0.5715742707252502,
+      "learning_rate": 1.3918464543538225e-06,
+      "loss": 0.08984670042991638,
+      "mean_token_accuracy": 0.9729854613542557,
+      "num_tokens": 9316805.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.205066344993968,
+      "eval_entropy": 0.2777416408899125,
+      "eval_loss": 0.9808777570724487,
+      "eval_mean_token_accuracy": 0.8252255926641185,
+      "eval_num_tokens": 9316805.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3820
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.932807665660928e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..0fadb03f84d5ddf3321159409a4aab2a85c47d4a
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/trainer_state.json
@@ -0,0 +1,4066 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.253317249698432,
+  "eval_steps": 20,
+  "global_step": 3840,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    },
+    {
+      "entropy": 0.17359357613783616,
+      "epoch": 9.012062726176115,
+      "grad_norm": 0.37213996052742004,
+      "learning_rate": 2.138437593000977e-06,
+      "loss": 0.10089871883392335,
+      "mean_token_accuracy": 0.9696638874518566,
+      "num_tokens": 9122597.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.012062726176115,
+      "eval_entropy": 0.2875421148505104,
+      "eval_loss": 0.9357590079307556,
+      "eval_mean_token_accuracy": 0.8264994859025719,
+      "eval_num_tokens": 9122597.0,
+      "eval_runtime": 98.523,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3740
+    },
+    {
+      "entropy": 0.1642639346420765,
+      "epoch": 9.060313630880579,
+      "grad_norm": 0.5511618852615356,
+      "learning_rate": 1.937215494938237e-06,
+      "loss": 0.08524587750434875,
+      "mean_token_accuracy": 0.9754035726189614,
+      "num_tokens": 9171477.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.060313630880579,
+      "eval_entropy": 0.2820046850469675,
+      "eval_loss": 0.9588018655776978,
+      "eval_mean_token_accuracy": 0.8257443647036392,
+      "eval_num_tokens": 9171477.0,
+      "eval_runtime": 98.3992,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1630103051662445,
+      "epoch": 9.108564535585042,
+      "grad_norm": 0.5125291347503662,
+      "learning_rate": 1.7456739289470143e-06,
+      "loss": 0.0895907461643219,
+      "mean_token_accuracy": 0.973100857436657,
+      "num_tokens": 9217543.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.108564535585042,
+      "eval_entropy": 0.2786683516388529,
+      "eval_loss": 0.9743365049362183,
+      "eval_mean_token_accuracy": 0.8253507131940863,
+      "eval_num_tokens": 9217543.0,
+      "eval_runtime": 98.3882,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3780
+    },
+    {
+      "entropy": 0.15873534101992845,
+      "epoch": 9.156815440289506,
+      "grad_norm": 0.5697775483131409,
+      "learning_rate": 1.5638670990247788e-06,
+      "loss": 0.08552997708320617,
+      "mean_token_accuracy": 0.973672679066658,
+      "num_tokens": 9268724.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.156815440289506,
+      "eval_entropy": 0.2781428360704626,
+      "eval_loss": 0.9790021777153015,
+      "eval_mean_token_accuracy": 0.8251898780967413,
+      "eval_num_tokens": 9268724.0,
+      "eval_runtime": 98.5449,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 3800
+    },
+    {
+      "entropy": 0.1606229657307267,
+      "epoch": 9.205066344993968,
+      "grad_norm": 0.5715742707252502,
+      "learning_rate": 1.3918464543538225e-06,
+      "loss": 0.08984670042991638,
+      "mean_token_accuracy": 0.9729854613542557,
+      "num_tokens": 9316805.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.205066344993968,
+      "eval_entropy": 0.2777416408899125,
+      "eval_loss": 0.9808777570724487,
+      "eval_mean_token_accuracy": 0.8252255926641185,
+      "eval_num_tokens": 9316805.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3820
+    },
+    {
+      "entropy": 0.15027269292622805,
+      "epoch": 9.253317249698432,
+      "grad_norm": 0.3359828293323517,
+      "learning_rate": 1.229660674741786e-06,
+      "loss": 0.08229721784591675,
+      "mean_token_accuracy": 0.9754723474383354,
+      "num_tokens": 9369940.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.253317249698432,
+      "eval_entropy": 0.2768587637483404,
+      "eval_loss": 0.9837455749511719,
+      "eval_mean_token_accuracy": 0.825493957889214,
+      "eval_num_tokens": 9369940.0,
+      "eval_runtime": 98.534,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 3840
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.943333795902464e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..dab472fdffda40c7d0f31c451315121788b999b8
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/trainer_state.json
@@ -0,0 +1,4087 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.301568154402895,
+  "eval_steps": 20,
+  "global_step": 3860,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    },
+    {
+      "entropy": 0.17359357613783616,
+      "epoch": 9.012062726176115,
+      "grad_norm": 0.37213996052742004,
+      "learning_rate": 2.138437593000977e-06,
+      "loss": 0.10089871883392335,
+      "mean_token_accuracy": 0.9696638874518566,
+      "num_tokens": 9122597.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.012062726176115,
+      "eval_entropy": 0.2875421148505104,
+      "eval_loss": 0.9357590079307556,
+      "eval_mean_token_accuracy": 0.8264994859025719,
+      "eval_num_tokens": 9122597.0,
+      "eval_runtime": 98.523,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3740
+    },
+    {
+      "entropy": 0.1642639346420765,
+      "epoch": 9.060313630880579,
+      "grad_norm": 0.5511618852615356,
+      "learning_rate": 1.937215494938237e-06,
+      "loss": 0.08524587750434875,
+      "mean_token_accuracy": 0.9754035726189614,
+      "num_tokens": 9171477.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.060313630880579,
+      "eval_entropy": 0.2820046850469675,
+      "eval_loss": 0.9588018655776978,
+      "eval_mean_token_accuracy": 0.8257443647036392,
+      "eval_num_tokens": 9171477.0,
+      "eval_runtime": 98.3992,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1630103051662445,
+      "epoch": 9.108564535585042,
+      "grad_norm": 0.5125291347503662,
+      "learning_rate": 1.7456739289470143e-06,
+      "loss": 0.0895907461643219,
+      "mean_token_accuracy": 0.973100857436657,
+      "num_tokens": 9217543.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.108564535585042,
+      "eval_entropy": 0.2786683516388529,
+      "eval_loss": 0.9743365049362183,
+      "eval_mean_token_accuracy": 0.8253507131940863,
+      "eval_num_tokens": 9217543.0,
+      "eval_runtime": 98.3882,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3780
+    },
+    {
+      "entropy": 0.15873534101992845,
+      "epoch": 9.156815440289506,
+      "grad_norm": 0.5697775483131409,
+      "learning_rate": 1.5638670990247788e-06,
+      "loss": 0.08552997708320617,
+      "mean_token_accuracy": 0.973672679066658,
+      "num_tokens": 9268724.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.156815440289506,
+      "eval_entropy": 0.2781428360704626,
+      "eval_loss": 0.9790021777153015,
+      "eval_mean_token_accuracy": 0.8251898780967413,
+      "eval_num_tokens": 9268724.0,
+      "eval_runtime": 98.5449,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 3800
+    },
+    {
+      "entropy": 0.1606229657307267,
+      "epoch": 9.205066344993968,
+      "grad_norm": 0.5715742707252502,
+      "learning_rate": 1.3918464543538225e-06,
+      "loss": 0.08984670042991638,
+      "mean_token_accuracy": 0.9729854613542557,
+      "num_tokens": 9316805.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.205066344993968,
+      "eval_entropy": 0.2777416408899125,
+      "eval_loss": 0.9808777570724487,
+      "eval_mean_token_accuracy": 0.8252255926641185,
+      "eval_num_tokens": 9316805.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3820
+    },
+    {
+      "entropy": 0.15027269292622805,
+      "epoch": 9.253317249698432,
+      "grad_norm": 0.3359828293323517,
+      "learning_rate": 1.229660674741786e-06,
+      "loss": 0.08229721784591675,
+      "mean_token_accuracy": 0.9754723474383354,
+      "num_tokens": 9369940.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.253317249698432,
+      "eval_entropy": 0.2768587637483404,
+      "eval_loss": 0.9837455749511719,
+      "eval_mean_token_accuracy": 0.825493957889214,
+      "eval_num_tokens": 9369940.0,
+      "eval_runtime": 98.534,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 3840
+    },
+    {
+      "entropy": 0.16767099779099226,
+      "epoch": 9.301568154402895,
+      "grad_norm": 0.5708875060081482,
+      "learning_rate": 1.077355656845822e-06,
+      "loss": 0.08888025283813476,
+      "mean_token_accuracy": 0.9715678468346596,
+      "num_tokens": 9416788.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.301568154402895,
+      "eval_entropy": 0.27612990283229377,
+      "eval_loss": 0.9873999357223511,
+      "eval_mean_token_accuracy": 0.8252144948150335,
+      "eval_num_tokens": 9416788.0,
+      "eval_runtime": 98.3889,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3860
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.953259488101376e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e614a101fe2410ef549408040c2f75d6e868d5b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/trainer_state.json
@@ -0,0 +1,4108 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.349819059107359,
+  "eval_steps": 20,
+  "global_step": 3880,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    },
+    {
+      "entropy": 0.17359357613783616,
+      "epoch": 9.012062726176115,
+      "grad_norm": 0.37213996052742004,
+      "learning_rate": 2.138437593000977e-06,
+      "loss": 0.10089871883392335,
+      "mean_token_accuracy": 0.9696638874518566,
+      "num_tokens": 9122597.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.012062726176115,
+      "eval_entropy": 0.2875421148505104,
+      "eval_loss": 0.9357590079307556,
+      "eval_mean_token_accuracy": 0.8264994859025719,
+      "eval_num_tokens": 9122597.0,
+      "eval_runtime": 98.523,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3740
+    },
+    {
+      "entropy": 0.1642639346420765,
+      "epoch": 9.060313630880579,
+      "grad_norm": 0.5511618852615356,
+      "learning_rate": 1.937215494938237e-06,
+      "loss": 0.08524587750434875,
+      "mean_token_accuracy": 0.9754035726189614,
+      "num_tokens": 9171477.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.060313630880579,
+      "eval_entropy": 0.2820046850469675,
+      "eval_loss": 0.9588018655776978,
+      "eval_mean_token_accuracy": 0.8257443647036392,
+      "eval_num_tokens": 9171477.0,
+      "eval_runtime": 98.3992,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1630103051662445,
+      "epoch": 9.108564535585042,
+      "grad_norm": 0.5125291347503662,
+      "learning_rate": 1.7456739289470143e-06,
+      "loss": 0.0895907461643219,
+      "mean_token_accuracy": 0.973100857436657,
+      "num_tokens": 9217543.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.108564535585042,
+      "eval_entropy": 0.2786683516388529,
+      "eval_loss": 0.9743365049362183,
+      "eval_mean_token_accuracy": 0.8253507131940863,
+      "eval_num_tokens": 9217543.0,
+      "eval_runtime": 98.3882,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3780
+    },
+    {
+      "entropy": 0.15873534101992845,
+      "epoch": 9.156815440289506,
+      "grad_norm": 0.5697775483131409,
+      "learning_rate": 1.5638670990247788e-06,
+      "loss": 0.08552997708320617,
+      "mean_token_accuracy": 0.973672679066658,
+      "num_tokens": 9268724.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.156815440289506,
+      "eval_entropy": 0.2781428360704626,
+      "eval_loss": 0.9790021777153015,
+      "eval_mean_token_accuracy": 0.8251898780967413,
+      "eval_num_tokens": 9268724.0,
+      "eval_runtime": 98.5449,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 3800
+    },
+    {
+      "entropy": 0.1606229657307267,
+      "epoch": 9.205066344993968,
+      "grad_norm": 0.5715742707252502,
+      "learning_rate": 1.3918464543538225e-06,
+      "loss": 0.08984670042991638,
+      "mean_token_accuracy": 0.9729854613542557,
+      "num_tokens": 9316805.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.205066344993968,
+      "eval_entropy": 0.2777416408899125,
+      "eval_loss": 0.9808777570724487,
+      "eval_mean_token_accuracy": 0.8252255926641185,
+      "eval_num_tokens": 9316805.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3820
+    },
+    {
+      "entropy": 0.15027269292622805,
+      "epoch": 9.253317249698432,
+      "grad_norm": 0.3359828293323517,
+      "learning_rate": 1.229660674741786e-06,
+      "loss": 0.08229721784591675,
+      "mean_token_accuracy": 0.9754723474383354,
+      "num_tokens": 9369940.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.253317249698432,
+      "eval_entropy": 0.2768587637483404,
+      "eval_loss": 0.9837455749511719,
+      "eval_mean_token_accuracy": 0.825493957889214,
+      "eval_num_tokens": 9369940.0,
+      "eval_runtime": 98.534,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 3840
+    },
+    {
+      "entropy": 0.16767099779099226,
+      "epoch": 9.301568154402895,
+      "grad_norm": 0.5708875060081482,
+      "learning_rate": 1.077355656845822e-06,
+      "loss": 0.08888025283813476,
+      "mean_token_accuracy": 0.9715678468346596,
+      "num_tokens": 9416788.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.301568154402895,
+      "eval_entropy": 0.27612990283229377,
+      "eval_loss": 0.9873999357223511,
+      "eval_mean_token_accuracy": 0.8252144948150335,
+      "eval_num_tokens": 9416788.0,
+      "eval_runtime": 98.3889,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3860
+    },
+    {
+      "entropy": 0.1549418892711401,
+      "epoch": 9.349819059107359,
+      "grad_norm": 0.7601456046104431,
+      "learning_rate": 9.349745011844029e-07,
+      "loss": 0.08411069512367249,
+      "mean_token_accuracy": 0.9741103485226631,
+      "num_tokens": 9467862.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.349819059107359,
+      "eval_entropy": 0.2757877716689967,
+      "eval_loss": 0.9892058968544006,
+      "eval_mean_token_accuracy": 0.8250612017143978,
+      "eval_num_tokens": 9467862.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3880
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.963460779994112e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..bab209f7caee98748fe8a8caffc40c938bb43596
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/trainer_state.json
@@ -0,0 +1,4129 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.398069963811821,
+  "eval_steps": 20,
+  "global_step": 3900,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    },
+    {
+      "entropy": 0.17359357613783616,
+      "epoch": 9.012062726176115,
+      "grad_norm": 0.37213996052742004,
+      "learning_rate": 2.138437593000977e-06,
+      "loss": 0.10089871883392335,
+      "mean_token_accuracy": 0.9696638874518566,
+      "num_tokens": 9122597.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.012062726176115,
+      "eval_entropy": 0.2875421148505104,
+      "eval_loss": 0.9357590079307556,
+      "eval_mean_token_accuracy": 0.8264994859025719,
+      "eval_num_tokens": 9122597.0,
+      "eval_runtime": 98.523,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3740
+    },
+    {
+      "entropy": 0.1642639346420765,
+      "epoch": 9.060313630880579,
+      "grad_norm": 0.5511618852615356,
+      "learning_rate": 1.937215494938237e-06,
+      "loss": 0.08524587750434875,
+      "mean_token_accuracy": 0.9754035726189614,
+      "num_tokens": 9171477.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.060313630880579,
+      "eval_entropy": 0.2820046850469675,
+      "eval_loss": 0.9588018655776978,
+      "eval_mean_token_accuracy": 0.8257443647036392,
+      "eval_num_tokens": 9171477.0,
+      "eval_runtime": 98.3992,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1630103051662445,
+      "epoch": 9.108564535585042,
+      "grad_norm": 0.5125291347503662,
+      "learning_rate": 1.7456739289470143e-06,
+      "loss": 0.0895907461643219,
+      "mean_token_accuracy": 0.973100857436657,
+      "num_tokens": 9217543.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.108564535585042,
+      "eval_entropy": 0.2786683516388529,
+      "eval_loss": 0.9743365049362183,
+      "eval_mean_token_accuracy": 0.8253507131940863,
+      "eval_num_tokens": 9217543.0,
+      "eval_runtime": 98.3882,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3780
+    },
+    {
+      "entropy": 0.15873534101992845,
+      "epoch": 9.156815440289506,
+      "grad_norm": 0.5697775483131409,
+      "learning_rate": 1.5638670990247788e-06,
+      "loss": 0.08552997708320617,
+      "mean_token_accuracy": 0.973672679066658,
+      "num_tokens": 9268724.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.156815440289506,
+      "eval_entropy": 0.2781428360704626,
+      "eval_loss": 0.9790021777153015,
+      "eval_mean_token_accuracy": 0.8251898780967413,
+      "eval_num_tokens": 9268724.0,
+      "eval_runtime": 98.5449,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 3800
+    },
+    {
+      "entropy": 0.1606229657307267,
+      "epoch": 9.205066344993968,
+      "grad_norm": 0.5715742707252502,
+      "learning_rate": 1.3918464543538225e-06,
+      "loss": 0.08984670042991638,
+      "mean_token_accuracy": 0.9729854613542557,
+      "num_tokens": 9316805.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.205066344993968,
+      "eval_entropy": 0.2777416408899125,
+      "eval_loss": 0.9808777570724487,
+      "eval_mean_token_accuracy": 0.8252255926641185,
+      "eval_num_tokens": 9316805.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3820
+    },
+    {
+      "entropy": 0.15027269292622805,
+      "epoch": 9.253317249698432,
+      "grad_norm": 0.3359828293323517,
+      "learning_rate": 1.229660674741786e-06,
+      "loss": 0.08229721784591675,
+      "mean_token_accuracy": 0.9754723474383354,
+      "num_tokens": 9369940.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.253317249698432,
+      "eval_entropy": 0.2768587637483404,
+      "eval_loss": 0.9837455749511719,
+      "eval_mean_token_accuracy": 0.825493957889214,
+      "eval_num_tokens": 9369940.0,
+      "eval_runtime": 98.534,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 3840
+    },
+    {
+      "entropy": 0.16767099779099226,
+      "epoch": 9.301568154402895,
+      "grad_norm": 0.5708875060081482,
+      "learning_rate": 1.077355656845822e-06,
+      "loss": 0.08888025283813476,
+      "mean_token_accuracy": 0.9715678468346596,
+      "num_tokens": 9416788.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.301568154402895,
+      "eval_entropy": 0.27612990283229377,
+      "eval_loss": 0.9873999357223511,
+      "eval_mean_token_accuracy": 0.8252144948150335,
+      "eval_num_tokens": 9416788.0,
+      "eval_runtime": 98.3889,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3860
+    },
+    {
+      "entropy": 0.1549418892711401,
+      "epoch": 9.349819059107359,
+      "grad_norm": 0.7601456046104431,
+      "learning_rate": 9.349745011844029e-07,
+      "loss": 0.08411069512367249,
+      "mean_token_accuracy": 0.9741103485226631,
+      "num_tokens": 9467862.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.349819059107359,
+      "eval_entropy": 0.2757877716689967,
+      "eval_loss": 0.9892058968544006,
+      "eval_mean_token_accuracy": 0.8250612017143978,
+      "eval_num_tokens": 9467862.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3880
+    },
+    {
+      "entropy": 0.16111517809331416,
+      "epoch": 9.398069963811821,
+      "grad_norm": 0.6774276494979858,
+      "learning_rate": 8.025574999403695e-07,
+      "loss": 0.08765592575073242,
+      "mean_token_accuracy": 0.9734174907207489,
+      "num_tokens": 9514611.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.398069963811821,
+      "eval_entropy": 0.275664389049739,
+      "eval_loss": 0.989257276058197,
+      "eval_mean_token_accuracy": 0.8252896914321385,
+      "eval_num_tokens": 9514611.0,
+      "eval_runtime": 98.4649,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 3900
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.97314848536064e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..0a53d60bd564caa8caf72aac3c387e833b762530
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/trainer_state.json
@@ -0,0 +1,4150 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.446320868516285,
+  "eval_steps": 20,
+  "global_step": 3920,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    },
+    {
+      "entropy": 0.17359357613783616,
+      "epoch": 9.012062726176115,
+      "grad_norm": 0.37213996052742004,
+      "learning_rate": 2.138437593000977e-06,
+      "loss": 0.10089871883392335,
+      "mean_token_accuracy": 0.9696638874518566,
+      "num_tokens": 9122597.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.012062726176115,
+      "eval_entropy": 0.2875421148505104,
+      "eval_loss": 0.9357590079307556,
+      "eval_mean_token_accuracy": 0.8264994859025719,
+      "eval_num_tokens": 9122597.0,
+      "eval_runtime": 98.523,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3740
+    },
+    {
+      "entropy": 0.1642639346420765,
+      "epoch": 9.060313630880579,
+      "grad_norm": 0.5511618852615356,
+      "learning_rate": 1.937215494938237e-06,
+      "loss": 0.08524587750434875,
+      "mean_token_accuracy": 0.9754035726189614,
+      "num_tokens": 9171477.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.060313630880579,
+      "eval_entropy": 0.2820046850469675,
+      "eval_loss": 0.9588018655776978,
+      "eval_mean_token_accuracy": 0.8257443647036392,
+      "eval_num_tokens": 9171477.0,
+      "eval_runtime": 98.3992,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1630103051662445,
+      "epoch": 9.108564535585042,
+      "grad_norm": 0.5125291347503662,
+      "learning_rate": 1.7456739289470143e-06,
+      "loss": 0.0895907461643219,
+      "mean_token_accuracy": 0.973100857436657,
+      "num_tokens": 9217543.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.108564535585042,
+      "eval_entropy": 0.2786683516388529,
+      "eval_loss": 0.9743365049362183,
+      "eval_mean_token_accuracy": 0.8253507131940863,
+      "eval_num_tokens": 9217543.0,
+      "eval_runtime": 98.3882,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3780
+    },
+    {
+      "entropy": 0.15873534101992845,
+      "epoch": 9.156815440289506,
+      "grad_norm": 0.5697775483131409,
+      "learning_rate": 1.5638670990247788e-06,
+      "loss": 0.08552997708320617,
+      "mean_token_accuracy": 0.973672679066658,
+      "num_tokens": 9268724.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.156815440289506,
+      "eval_entropy": 0.2781428360704626,
+      "eval_loss": 0.9790021777153015,
+      "eval_mean_token_accuracy": 0.8251898780967413,
+      "eval_num_tokens": 9268724.0,
+      "eval_runtime": 98.5449,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 3800
+    },
+    {
+      "entropy": 0.1606229657307267,
+      "epoch": 9.205066344993968,
+      "grad_norm": 0.5715742707252502,
+      "learning_rate": 1.3918464543538225e-06,
+      "loss": 0.08984670042991638,
+      "mean_token_accuracy": 0.9729854613542557,
+      "num_tokens": 9316805.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.205066344993968,
+      "eval_entropy": 0.2777416408899125,
+      "eval_loss": 0.9808777570724487,
+      "eval_mean_token_accuracy": 0.8252255926641185,
+      "eval_num_tokens": 9316805.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3820
+    },
+    {
+      "entropy": 0.15027269292622805,
+      "epoch": 9.253317249698432,
+      "grad_norm": 0.3359828293323517,
+      "learning_rate": 1.229660674741786e-06,
+      "loss": 0.08229721784591675,
+      "mean_token_accuracy": 0.9754723474383354,
+      "num_tokens": 9369940.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.253317249698432,
+      "eval_entropy": 0.2768587637483404,
+      "eval_loss": 0.9837455749511719,
+      "eval_mean_token_accuracy": 0.825493957889214,
+      "eval_num_tokens": 9369940.0,
+      "eval_runtime": 98.534,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 3840
+    },
+    {
+      "entropy": 0.16767099779099226,
+      "epoch": 9.301568154402895,
+      "grad_norm": 0.5708875060081482,
+      "learning_rate": 1.077355656845822e-06,
+      "loss": 0.08888025283813476,
+      "mean_token_accuracy": 0.9715678468346596,
+      "num_tokens": 9416788.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.301568154402895,
+      "eval_entropy": 0.27612990283229377,
+      "eval_loss": 0.9873999357223511,
+      "eval_mean_token_accuracy": 0.8252144948150335,
+      "eval_num_tokens": 9416788.0,
+      "eval_runtime": 98.3889,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3860
+    },
+    {
+      "entropy": 0.1549418892711401,
+      "epoch": 9.349819059107359,
+      "grad_norm": 0.7601456046104431,
+      "learning_rate": 9.349745011844029e-07,
+      "loss": 0.08411069512367249,
+      "mean_token_accuracy": 0.9741103485226631,
+      "num_tokens": 9467862.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.349819059107359,
+      "eval_entropy": 0.2757877716689967,
+      "eval_loss": 0.9892058968544006,
+      "eval_mean_token_accuracy": 0.8250612017143978,
+      "eval_num_tokens": 9467862.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3880
+    },
+    {
+      "entropy": 0.16111517809331416,
+      "epoch": 9.398069963811821,
+      "grad_norm": 0.6774276494979858,
+      "learning_rate": 8.025574999403695e-07,
+      "loss": 0.08765592575073242,
+      "mean_token_accuracy": 0.9734174907207489,
+      "num_tokens": 9514611.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.398069963811821,
+      "eval_entropy": 0.275664389049739,
+      "eval_loss": 0.989257276058197,
+      "eval_mean_token_accuracy": 0.8252896914321385,
+      "eval_num_tokens": 9514611.0,
+      "eval_runtime": 98.4649,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 3900
+    },
+    {
+      "entropy": 0.15701593235135078,
+      "epoch": 9.446320868516285,
+      "grad_norm": 0.6130298376083374,
+      "learning_rate": 6.801421255587377e-07,
+      "loss": 0.0853091835975647,
+      "mean_token_accuracy": 0.9740088582038879,
+      "num_tokens": 9563437.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.446320868516285,
+      "eval_entropy": 0.27559053881114787,
+      "eval_loss": 0.9901270866394043,
+      "eval_mean_token_accuracy": 0.8251507653948966,
+      "eval_num_tokens": 9563437.0,
+      "eval_runtime": 98.5285,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 3920
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.983131622657024e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c5e70c6cafc4f2819af5d0ff776f07b42075bf05
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/trainer_state.json
@@ -0,0 +1,4171 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.494571773220748,
+  "eval_steps": 20,
+  "global_step": 3940,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    },
+    {
+      "entropy": 0.17359357613783616,
+      "epoch": 9.012062726176115,
+      "grad_norm": 0.37213996052742004,
+      "learning_rate": 2.138437593000977e-06,
+      "loss": 0.10089871883392335,
+      "mean_token_accuracy": 0.9696638874518566,
+      "num_tokens": 9122597.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.012062726176115,
+      "eval_entropy": 0.2875421148505104,
+      "eval_loss": 0.9357590079307556,
+      "eval_mean_token_accuracy": 0.8264994859025719,
+      "eval_num_tokens": 9122597.0,
+      "eval_runtime": 98.523,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3740
+    },
+    {
+      "entropy": 0.1642639346420765,
+      "epoch": 9.060313630880579,
+      "grad_norm": 0.5511618852615356,
+      "learning_rate": 1.937215494938237e-06,
+      "loss": 0.08524587750434875,
+      "mean_token_accuracy": 0.9754035726189614,
+      "num_tokens": 9171477.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.060313630880579,
+      "eval_entropy": 0.2820046850469675,
+      "eval_loss": 0.9588018655776978,
+      "eval_mean_token_accuracy": 0.8257443647036392,
+      "eval_num_tokens": 9171477.0,
+      "eval_runtime": 98.3992,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1630103051662445,
+      "epoch": 9.108564535585042,
+      "grad_norm": 0.5125291347503662,
+      "learning_rate": 1.7456739289470143e-06,
+      "loss": 0.0895907461643219,
+      "mean_token_accuracy": 0.973100857436657,
+      "num_tokens": 9217543.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.108564535585042,
+      "eval_entropy": 0.2786683516388529,
+      "eval_loss": 0.9743365049362183,
+      "eval_mean_token_accuracy": 0.8253507131940863,
+      "eval_num_tokens": 9217543.0,
+      "eval_runtime": 98.3882,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3780
+    },
+    {
+      "entropy": 0.15873534101992845,
+      "epoch": 9.156815440289506,
+      "grad_norm": 0.5697775483131409,
+      "learning_rate": 1.5638670990247788e-06,
+      "loss": 0.08552997708320617,
+      "mean_token_accuracy": 0.973672679066658,
+      "num_tokens": 9268724.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.156815440289506,
+      "eval_entropy": 0.2781428360704626,
+      "eval_loss": 0.9790021777153015,
+      "eval_mean_token_accuracy": 0.8251898780967413,
+      "eval_num_tokens": 9268724.0,
+      "eval_runtime": 98.5449,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 3800
+    },
+    {
+      "entropy": 0.1606229657307267,
+      "epoch": 9.205066344993968,
+      "grad_norm": 0.5715742707252502,
+      "learning_rate": 1.3918464543538225e-06,
+      "loss": 0.08984670042991638,
+      "mean_token_accuracy": 0.9729854613542557,
+      "num_tokens": 9316805.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.205066344993968,
+      "eval_entropy": 0.2777416408899125,
+      "eval_loss": 0.9808777570724487,
+      "eval_mean_token_accuracy": 0.8252255926641185,
+      "eval_num_tokens": 9316805.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3820
+    },
+    {
+      "entropy": 0.15027269292622805,
+      "epoch": 9.253317249698432,
+      "grad_norm": 0.3359828293323517,
+      "learning_rate": 1.229660674741786e-06,
+      "loss": 0.08229721784591675,
+      "mean_token_accuracy": 0.9754723474383354,
+      "num_tokens": 9369940.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.253317249698432,
+      "eval_entropy": 0.2768587637483404,
+      "eval_loss": 0.9837455749511719,
+      "eval_mean_token_accuracy": 0.825493957889214,
+      "eval_num_tokens": 9369940.0,
+      "eval_runtime": 98.534,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 3840
+    },
+    {
+      "entropy": 0.16767099779099226,
+      "epoch": 9.301568154402895,
+      "grad_norm": 0.5708875060081482,
+      "learning_rate": 1.077355656845822e-06,
+      "loss": 0.08888025283813476,
+      "mean_token_accuracy": 0.9715678468346596,
+      "num_tokens": 9416788.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.301568154402895,
+      "eval_entropy": 0.27612990283229377,
+      "eval_loss": 0.9873999357223511,
+      "eval_mean_token_accuracy": 0.8252144948150335,
+      "eval_num_tokens": 9416788.0,
+      "eval_runtime": 98.3889,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3860
+    },
+    {
+      "entropy": 0.1549418892711401,
+      "epoch": 9.349819059107359,
+      "grad_norm": 0.7601456046104431,
+      "learning_rate": 9.349745011844029e-07,
+      "loss": 0.08411069512367249,
+      "mean_token_accuracy": 0.9741103485226631,
+      "num_tokens": 9467862.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.349819059107359,
+      "eval_entropy": 0.2757877716689967,
+      "eval_loss": 0.9892058968544006,
+      "eval_mean_token_accuracy": 0.8250612017143978,
+      "eval_num_tokens": 9467862.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3880
+    },
+    {
+      "entropy": 0.16111517809331416,
+      "epoch": 9.398069963811821,
+      "grad_norm": 0.6774276494979858,
+      "learning_rate": 8.025574999403695e-07,
+      "loss": 0.08765592575073242,
+      "mean_token_accuracy": 0.9734174907207489,
+      "num_tokens": 9514611.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.398069963811821,
+      "eval_entropy": 0.275664389049739,
+      "eval_loss": 0.989257276058197,
+      "eval_mean_token_accuracy": 0.8252896914321385,
+      "eval_num_tokens": 9514611.0,
+      "eval_runtime": 98.4649,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 3900
+    },
+    {
+      "entropy": 0.15701593235135078,
+      "epoch": 9.446320868516285,
+      "grad_norm": 0.6130298376083374,
+      "learning_rate": 6.801421255587377e-07,
+      "loss": 0.0853091835975647,
+      "mean_token_accuracy": 0.9740088582038879,
+      "num_tokens": 9563437.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.446320868516285,
+      "eval_entropy": 0.27559053881114787,
+      "eval_loss": 0.9901270866394043,
+      "eval_mean_token_accuracy": 0.8251507653948966,
+      "eval_num_tokens": 9563437.0,
+      "eval_runtime": 98.5285,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 3920
+    },
+    {
+      "entropy": 0.15848954878747462,
+      "epoch": 9.494571773220748,
+      "grad_norm": 0.5928868651390076,
+      "learning_rate": 5.677630201424204e-07,
+      "loss": 0.08467000126838684,
+      "mean_token_accuracy": 0.9725634664297104,
+      "num_tokens": 9614034.0,
+      "step": 3940
+    },
+    {
+      "epoch": 9.494571773220748,
+      "eval_entropy": 0.2750965692521481,
+      "eval_loss": 0.9914345145225525,
+      "eval_mean_token_accuracy": 0.8252089954494091,
+      "eval_num_tokens": 9614034.0,
+      "eval_runtime": 98.5049,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3940
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.99420484906496e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..836123554f9bd48eaaee378a52b600d72d39a2fd
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/trainer_state.json
@@ -0,0 +1,4192 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.54282267792521,
+  "eval_steps": 20,
+  "global_step": 3960,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    },
+    {
+      "entropy": 0.17359357613783616,
+      "epoch": 9.012062726176115,
+      "grad_norm": 0.37213996052742004,
+      "learning_rate": 2.138437593000977e-06,
+      "loss": 0.10089871883392335,
+      "mean_token_accuracy": 0.9696638874518566,
+      "num_tokens": 9122597.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.012062726176115,
+      "eval_entropy": 0.2875421148505104,
+      "eval_loss": 0.9357590079307556,
+      "eval_mean_token_accuracy": 0.8264994859025719,
+      "eval_num_tokens": 9122597.0,
+      "eval_runtime": 98.523,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3740
+    },
+    {
+      "entropy": 0.1642639346420765,
+      "epoch": 9.060313630880579,
+      "grad_norm": 0.5511618852615356,
+      "learning_rate": 1.937215494938237e-06,
+      "loss": 0.08524587750434875,
+      "mean_token_accuracy": 0.9754035726189614,
+      "num_tokens": 9171477.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.060313630880579,
+      "eval_entropy": 0.2820046850469675,
+      "eval_loss": 0.9588018655776978,
+      "eval_mean_token_accuracy": 0.8257443647036392,
+      "eval_num_tokens": 9171477.0,
+      "eval_runtime": 98.3992,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1630103051662445,
+      "epoch": 9.108564535585042,
+      "grad_norm": 0.5125291347503662,
+      "learning_rate": 1.7456739289470143e-06,
+      "loss": 0.0895907461643219,
+      "mean_token_accuracy": 0.973100857436657,
+      "num_tokens": 9217543.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.108564535585042,
+      "eval_entropy": 0.2786683516388529,
+      "eval_loss": 0.9743365049362183,
+      "eval_mean_token_accuracy": 0.8253507131940863,
+      "eval_num_tokens": 9217543.0,
+      "eval_runtime": 98.3882,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3780
+    },
+    {
+      "entropy": 0.15873534101992845,
+      "epoch": 9.156815440289506,
+      "grad_norm": 0.5697775483131409,
+      "learning_rate": 1.5638670990247788e-06,
+      "loss": 0.08552997708320617,
+      "mean_token_accuracy": 0.973672679066658,
+      "num_tokens": 9268724.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.156815440289506,
+      "eval_entropy": 0.2781428360704626,
+      "eval_loss": 0.9790021777153015,
+      "eval_mean_token_accuracy": 0.8251898780967413,
+      "eval_num_tokens": 9268724.0,
+      "eval_runtime": 98.5449,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 3800
+    },
+    {
+      "entropy": 0.1606229657307267,
+      "epoch": 9.205066344993968,
+      "grad_norm": 0.5715742707252502,
+      "learning_rate": 1.3918464543538225e-06,
+      "loss": 0.08984670042991638,
+      "mean_token_accuracy": 0.9729854613542557,
+      "num_tokens": 9316805.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.205066344993968,
+      "eval_entropy": 0.2777416408899125,
+      "eval_loss": 0.9808777570724487,
+      "eval_mean_token_accuracy": 0.8252255926641185,
+      "eval_num_tokens": 9316805.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3820
+    },
+    {
+      "entropy": 0.15027269292622805,
+      "epoch": 9.253317249698432,
+      "grad_norm": 0.3359828293323517,
+      "learning_rate": 1.229660674741786e-06,
+      "loss": 0.08229721784591675,
+      "mean_token_accuracy": 0.9754723474383354,
+      "num_tokens": 9369940.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.253317249698432,
+      "eval_entropy": 0.2768587637483404,
+      "eval_loss": 0.9837455749511719,
+      "eval_mean_token_accuracy": 0.825493957889214,
+      "eval_num_tokens": 9369940.0,
+      "eval_runtime": 98.534,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 3840
+    },
+    {
+      "entropy": 0.16767099779099226,
+      "epoch": 9.301568154402895,
+      "grad_norm": 0.5708875060081482,
+      "learning_rate": 1.077355656845822e-06,
+      "loss": 0.08888025283813476,
+      "mean_token_accuracy": 0.9715678468346596,
+      "num_tokens": 9416788.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.301568154402895,
+      "eval_entropy": 0.27612990283229377,
+      "eval_loss": 0.9873999357223511,
+      "eval_mean_token_accuracy": 0.8252144948150335,
+      "eval_num_tokens": 9416788.0,
+      "eval_runtime": 98.3889,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3860
+    },
+    {
+      "entropy": 0.1549418892711401,
+      "epoch": 9.349819059107359,
+      "grad_norm": 0.7601456046104431,
+      "learning_rate": 9.349745011844029e-07,
+      "loss": 0.08411069512367249,
+      "mean_token_accuracy": 0.9741103485226631,
+      "num_tokens": 9467862.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.349819059107359,
+      "eval_entropy": 0.2757877716689967,
+      "eval_loss": 0.9892058968544006,
+      "eval_mean_token_accuracy": 0.8250612017143978,
+      "eval_num_tokens": 9467862.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3880
+    },
+    {
+      "entropy": 0.16111517809331416,
+      "epoch": 9.398069963811821,
+      "grad_norm": 0.6774276494979858,
+      "learning_rate": 8.025574999403695e-07,
+      "loss": 0.08765592575073242,
+      "mean_token_accuracy": 0.9734174907207489,
+      "num_tokens": 9514611.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.398069963811821,
+      "eval_entropy": 0.275664389049739,
+      "eval_loss": 0.989257276058197,
+      "eval_mean_token_accuracy": 0.8252896914321385,
+      "eval_num_tokens": 9514611.0,
+      "eval_runtime": 98.4649,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 3900
+    },
+    {
+      "entropy": 0.15701593235135078,
+      "epoch": 9.446320868516285,
+      "grad_norm": 0.6130298376083374,
+      "learning_rate": 6.801421255587377e-07,
+      "loss": 0.0853091835975647,
+      "mean_token_accuracy": 0.9740088582038879,
+      "num_tokens": 9563437.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.446320868516285,
+      "eval_entropy": 0.27559053881114787,
+      "eval_loss": 0.9901270866394043,
+      "eval_mean_token_accuracy": 0.8251507653948966,
+      "eval_num_tokens": 9563437.0,
+      "eval_runtime": 98.5285,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 3920
+    },
+    {
+      "entropy": 0.15848954878747462,
+      "epoch": 9.494571773220748,
+      "grad_norm": 0.5928868651390076,
+      "learning_rate": 5.677630201424204e-07,
+      "loss": 0.08467000126838684,
+      "mean_token_accuracy": 0.9725634664297104,
+      "num_tokens": 9614034.0,
+      "step": 3940
+    },
+    {
+      "epoch": 9.494571773220748,
+      "eval_entropy": 0.2750965692521481,
+      "eval_loss": 0.9914345145225525,
+      "eval_mean_token_accuracy": 0.8252089954494091,
+      "eval_num_tokens": 9614034.0,
+      "eval_runtime": 98.5049,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3940
+    },
+    {
+      "entropy": 0.15644967313855887,
+      "epoch": 9.54282267792521,
+      "grad_norm": 0.5718610882759094,
+      "learning_rate": 4.6545198564893463e-07,
+      "loss": 0.08564762473106384,
+      "mean_token_accuracy": 0.9730244860053062,
+      "num_tokens": 9661266.0,
+      "step": 3960
+    },
+    {
+      "epoch": 9.54282267792521,
+      "eval_entropy": 0.27522430820076654,
+      "eval_loss": 0.9908939003944397,
+      "eval_mean_token_accuracy": 0.8251052994406625,
+      "eval_num_tokens": 9661266.0,
+      "eval_runtime": 98.3637,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3960
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.00357045156352e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e41b15665ccd9fcf398784e32a49458970c963ec
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/trainer_state.json
@@ -0,0 +1,4213 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.591073582629674,
+  "eval_steps": 20,
+  "global_step": 3980,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    },
+    {
+      "entropy": 0.17359357613783616,
+      "epoch": 9.012062726176115,
+      "grad_norm": 0.37213996052742004,
+      "learning_rate": 2.138437593000977e-06,
+      "loss": 0.10089871883392335,
+      "mean_token_accuracy": 0.9696638874518566,
+      "num_tokens": 9122597.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.012062726176115,
+      "eval_entropy": 0.2875421148505104,
+      "eval_loss": 0.9357590079307556,
+      "eval_mean_token_accuracy": 0.8264994859025719,
+      "eval_num_tokens": 9122597.0,
+      "eval_runtime": 98.523,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3740
+    },
+    {
+      "entropy": 0.1642639346420765,
+      "epoch": 9.060313630880579,
+      "grad_norm": 0.5511618852615356,
+      "learning_rate": 1.937215494938237e-06,
+      "loss": 0.08524587750434875,
+      "mean_token_accuracy": 0.9754035726189614,
+      "num_tokens": 9171477.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.060313630880579,
+      "eval_entropy": 0.2820046850469675,
+      "eval_loss": 0.9588018655776978,
+      "eval_mean_token_accuracy": 0.8257443647036392,
+      "eval_num_tokens": 9171477.0,
+      "eval_runtime": 98.3992,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1630103051662445,
+      "epoch": 9.108564535585042,
+      "grad_norm": 0.5125291347503662,
+      "learning_rate": 1.7456739289470143e-06,
+      "loss": 0.0895907461643219,
+      "mean_token_accuracy": 0.973100857436657,
+      "num_tokens": 9217543.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.108564535585042,
+      "eval_entropy": 0.2786683516388529,
+      "eval_loss": 0.9743365049362183,
+      "eval_mean_token_accuracy": 0.8253507131940863,
+      "eval_num_tokens": 9217543.0,
+      "eval_runtime": 98.3882,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3780
+    },
+    {
+      "entropy": 0.15873534101992845,
+      "epoch": 9.156815440289506,
+      "grad_norm": 0.5697775483131409,
+      "learning_rate": 1.5638670990247788e-06,
+      "loss": 0.08552997708320617,
+      "mean_token_accuracy": 0.973672679066658,
+      "num_tokens": 9268724.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.156815440289506,
+      "eval_entropy": 0.2781428360704626,
+      "eval_loss": 0.9790021777153015,
+      "eval_mean_token_accuracy": 0.8251898780967413,
+      "eval_num_tokens": 9268724.0,
+      "eval_runtime": 98.5449,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 3800
+    },
+    {
+      "entropy": 0.1606229657307267,
+      "epoch": 9.205066344993968,
+      "grad_norm": 0.5715742707252502,
+      "learning_rate": 1.3918464543538225e-06,
+      "loss": 0.08984670042991638,
+      "mean_token_accuracy": 0.9729854613542557,
+      "num_tokens": 9316805.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.205066344993968,
+      "eval_entropy": 0.2777416408899125,
+      "eval_loss": 0.9808777570724487,
+      "eval_mean_token_accuracy": 0.8252255926641185,
+      "eval_num_tokens": 9316805.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3820
+    },
+    {
+      "entropy": 0.15027269292622805,
+      "epoch": 9.253317249698432,
+      "grad_norm": 0.3359828293323517,
+      "learning_rate": 1.229660674741786e-06,
+      "loss": 0.08229721784591675,
+      "mean_token_accuracy": 0.9754723474383354,
+      "num_tokens": 9369940.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.253317249698432,
+      "eval_entropy": 0.2768587637483404,
+      "eval_loss": 0.9837455749511719,
+      "eval_mean_token_accuracy": 0.825493957889214,
+      "eval_num_tokens": 9369940.0,
+      "eval_runtime": 98.534,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 3840
+    },
+    {
+      "entropy": 0.16767099779099226,
+      "epoch": 9.301568154402895,
+      "grad_norm": 0.5708875060081482,
+      "learning_rate": 1.077355656845822e-06,
+      "loss": 0.08888025283813476,
+      "mean_token_accuracy": 0.9715678468346596,
+      "num_tokens": 9416788.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.301568154402895,
+      "eval_entropy": 0.27612990283229377,
+      "eval_loss": 0.9873999357223511,
+      "eval_mean_token_accuracy": 0.8252144948150335,
+      "eval_num_tokens": 9416788.0,
+      "eval_runtime": 98.3889,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3860
+    },
+    {
+      "entropy": 0.1549418892711401,
+      "epoch": 9.349819059107359,
+      "grad_norm": 0.7601456046104431,
+      "learning_rate": 9.349745011844029e-07,
+      "loss": 0.08411069512367249,
+      "mean_token_accuracy": 0.9741103485226631,
+      "num_tokens": 9467862.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.349819059107359,
+      "eval_entropy": 0.2757877716689967,
+      "eval_loss": 0.9892058968544006,
+      "eval_mean_token_accuracy": 0.8250612017143978,
+      "eval_num_tokens": 9467862.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3880
+    },
+    {
+      "entropy": 0.16111517809331416,
+      "epoch": 9.398069963811821,
+      "grad_norm": 0.6774276494979858,
+      "learning_rate": 8.025574999403695e-07,
+      "loss": 0.08765592575073242,
+      "mean_token_accuracy": 0.9734174907207489,
+      "num_tokens": 9514611.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.398069963811821,
+      "eval_entropy": 0.275664389049739,
+      "eval_loss": 0.989257276058197,
+      "eval_mean_token_accuracy": 0.8252896914321385,
+      "eval_num_tokens": 9514611.0,
+      "eval_runtime": 98.4649,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 3900
+    },
+    {
+      "entropy": 0.15701593235135078,
+      "epoch": 9.446320868516285,
+      "grad_norm": 0.6130298376083374,
+      "learning_rate": 6.801421255587377e-07,
+      "loss": 0.0853091835975647,
+      "mean_token_accuracy": 0.9740088582038879,
+      "num_tokens": 9563437.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.446320868516285,
+      "eval_entropy": 0.27559053881114787,
+      "eval_loss": 0.9901270866394043,
+      "eval_mean_token_accuracy": 0.8251507653948966,
+      "eval_num_tokens": 9563437.0,
+      "eval_runtime": 98.5285,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 3920
+    },
+    {
+      "entropy": 0.15848954878747462,
+      "epoch": 9.494571773220748,
+      "grad_norm": 0.5928868651390076,
+      "learning_rate": 5.677630201424204e-07,
+      "loss": 0.08467000126838684,
+      "mean_token_accuracy": 0.9725634664297104,
+      "num_tokens": 9614034.0,
+      "step": 3940
+    },
+    {
+      "epoch": 9.494571773220748,
+      "eval_entropy": 0.2750965692521481,
+      "eval_loss": 0.9914345145225525,
+      "eval_mean_token_accuracy": 0.8252089954494091,
+      "eval_num_tokens": 9614034.0,
+      "eval_runtime": 98.5049,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3940
+    },
+    {
+      "entropy": 0.15644967313855887,
+      "epoch": 9.54282267792521,
+      "grad_norm": 0.5718610882759094,
+      "learning_rate": 4.6545198564893463e-07,
+      "loss": 0.08564762473106384,
+      "mean_token_accuracy": 0.9730244860053062,
+      "num_tokens": 9661266.0,
+      "step": 3960
+    },
+    {
+      "epoch": 9.54282267792521,
+      "eval_entropy": 0.27522430820076654,
+      "eval_loss": 0.9908939003944397,
+      "eval_mean_token_accuracy": 0.8251052994406625,
+      "eval_num_tokens": 9661266.0,
+      "eval_runtime": 98.3637,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3960
+    },
+    {
+      "entropy": 0.15196608882397414,
+      "epoch": 9.591073582629674,
+      "grad_norm": 0.752661406993866,
+      "learning_rate": 3.732379748908161e-07,
+      "loss": 0.08072127103805542,
+      "mean_token_accuracy": 0.9749975100159645,
+      "num_tokens": 9710458.0,
+      "step": 3980
+    },
+    {
+      "epoch": 9.591073582629674,
+      "eval_entropy": 0.2750371985890892,
+      "eval_loss": 0.9913897514343262,
+      "eval_mean_token_accuracy": 0.8248539731073915,
+      "eval_num_tokens": 9710458.0,
+      "eval_runtime": 98.4959,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3980
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.012769189731328e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..5aea4d275b2c944fd0bdd987807b86fed3d2b360
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/trainer_state.json
@@ -0,0 +1,76 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.09650180940892641,
+  "eval_steps": 20,
+  "global_step": 40,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.9962855241728e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..257d06cf2ccb55953fbbb2f8c97315fdb9ddbe8a
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/trainer_state.json
@@ -0,0 +1,454 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9650180940892642,
+  "eval_steps": 20,
+  "global_step": 400,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.03802896166912e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..009a42df44bc622fe3ef9b1e8a0b0481f25782c9
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/trainer_state.json
@@ -0,0 +1,4234 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.639324487334138,
+  "eval_steps": 20,
+  "global_step": 4000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    },
+    {
+      "entropy": 0.17359357613783616,
+      "epoch": 9.012062726176115,
+      "grad_norm": 0.37213996052742004,
+      "learning_rate": 2.138437593000977e-06,
+      "loss": 0.10089871883392335,
+      "mean_token_accuracy": 0.9696638874518566,
+      "num_tokens": 9122597.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.012062726176115,
+      "eval_entropy": 0.2875421148505104,
+      "eval_loss": 0.9357590079307556,
+      "eval_mean_token_accuracy": 0.8264994859025719,
+      "eval_num_tokens": 9122597.0,
+      "eval_runtime": 98.523,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3740
+    },
+    {
+      "entropy": 0.1642639346420765,
+      "epoch": 9.060313630880579,
+      "grad_norm": 0.5511618852615356,
+      "learning_rate": 1.937215494938237e-06,
+      "loss": 0.08524587750434875,
+      "mean_token_accuracy": 0.9754035726189614,
+      "num_tokens": 9171477.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.060313630880579,
+      "eval_entropy": 0.2820046850469675,
+      "eval_loss": 0.9588018655776978,
+      "eval_mean_token_accuracy": 0.8257443647036392,
+      "eval_num_tokens": 9171477.0,
+      "eval_runtime": 98.3992,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1630103051662445,
+      "epoch": 9.108564535585042,
+      "grad_norm": 0.5125291347503662,
+      "learning_rate": 1.7456739289470143e-06,
+      "loss": 0.0895907461643219,
+      "mean_token_accuracy": 0.973100857436657,
+      "num_tokens": 9217543.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.108564535585042,
+      "eval_entropy": 0.2786683516388529,
+      "eval_loss": 0.9743365049362183,
+      "eval_mean_token_accuracy": 0.8253507131940863,
+      "eval_num_tokens": 9217543.0,
+      "eval_runtime": 98.3882,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3780
+    },
+    {
+      "entropy": 0.15873534101992845,
+      "epoch": 9.156815440289506,
+      "grad_norm": 0.5697775483131409,
+      "learning_rate": 1.5638670990247788e-06,
+      "loss": 0.08552997708320617,
+      "mean_token_accuracy": 0.973672679066658,
+      "num_tokens": 9268724.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.156815440289506,
+      "eval_entropy": 0.2781428360704626,
+      "eval_loss": 0.9790021777153015,
+      "eval_mean_token_accuracy": 0.8251898780967413,
+      "eval_num_tokens": 9268724.0,
+      "eval_runtime": 98.5449,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 3800
+    },
+    {
+      "entropy": 0.1606229657307267,
+      "epoch": 9.205066344993968,
+      "grad_norm": 0.5715742707252502,
+      "learning_rate": 1.3918464543538225e-06,
+      "loss": 0.08984670042991638,
+      "mean_token_accuracy": 0.9729854613542557,
+      "num_tokens": 9316805.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.205066344993968,
+      "eval_entropy": 0.2777416408899125,
+      "eval_loss": 0.9808777570724487,
+      "eval_mean_token_accuracy": 0.8252255926641185,
+      "eval_num_tokens": 9316805.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3820
+    },
+    {
+      "entropy": 0.15027269292622805,
+      "epoch": 9.253317249698432,
+      "grad_norm": 0.3359828293323517,
+      "learning_rate": 1.229660674741786e-06,
+      "loss": 0.08229721784591675,
+      "mean_token_accuracy": 0.9754723474383354,
+      "num_tokens": 9369940.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.253317249698432,
+      "eval_entropy": 0.2768587637483404,
+      "eval_loss": 0.9837455749511719,
+      "eval_mean_token_accuracy": 0.825493957889214,
+      "eval_num_tokens": 9369940.0,
+      "eval_runtime": 98.534,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 3840
+    },
+    {
+      "entropy": 0.16767099779099226,
+      "epoch": 9.301568154402895,
+      "grad_norm": 0.5708875060081482,
+      "learning_rate": 1.077355656845822e-06,
+      "loss": 0.08888025283813476,
+      "mean_token_accuracy": 0.9715678468346596,
+      "num_tokens": 9416788.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.301568154402895,
+      "eval_entropy": 0.27612990283229377,
+      "eval_loss": 0.9873999357223511,
+      "eval_mean_token_accuracy": 0.8252144948150335,
+      "eval_num_tokens": 9416788.0,
+      "eval_runtime": 98.3889,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3860
+    },
+    {
+      "entropy": 0.1549418892711401,
+      "epoch": 9.349819059107359,
+      "grad_norm": 0.7601456046104431,
+      "learning_rate": 9.349745011844029e-07,
+      "loss": 0.08411069512367249,
+      "mean_token_accuracy": 0.9741103485226631,
+      "num_tokens": 9467862.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.349819059107359,
+      "eval_entropy": 0.2757877716689967,
+      "eval_loss": 0.9892058968544006,
+      "eval_mean_token_accuracy": 0.8250612017143978,
+      "eval_num_tokens": 9467862.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3880
+    },
+    {
+      "entropy": 0.16111517809331416,
+      "epoch": 9.398069963811821,
+      "grad_norm": 0.6774276494979858,
+      "learning_rate": 8.025574999403695e-07,
+      "loss": 0.08765592575073242,
+      "mean_token_accuracy": 0.9734174907207489,
+      "num_tokens": 9514611.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.398069963811821,
+      "eval_entropy": 0.275664389049739,
+      "eval_loss": 0.989257276058197,
+      "eval_mean_token_accuracy": 0.8252896914321385,
+      "eval_num_tokens": 9514611.0,
+      "eval_runtime": 98.4649,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 3900
+    },
+    {
+      "entropy": 0.15701593235135078,
+      "epoch": 9.446320868516285,
+      "grad_norm": 0.6130298376083374,
+      "learning_rate": 6.801421255587377e-07,
+      "loss": 0.0853091835975647,
+      "mean_token_accuracy": 0.9740088582038879,
+      "num_tokens": 9563437.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.446320868516285,
+      "eval_entropy": 0.27559053881114787,
+      "eval_loss": 0.9901270866394043,
+      "eval_mean_token_accuracy": 0.8251507653948966,
+      "eval_num_tokens": 9563437.0,
+      "eval_runtime": 98.5285,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 3920
+    },
+    {
+      "entropy": 0.15848954878747462,
+      "epoch": 9.494571773220748,
+      "grad_norm": 0.5928868651390076,
+      "learning_rate": 5.677630201424204e-07,
+      "loss": 0.08467000126838684,
+      "mean_token_accuracy": 0.9725634664297104,
+      "num_tokens": 9614034.0,
+      "step": 3940
+    },
+    {
+      "epoch": 9.494571773220748,
+      "eval_entropy": 0.2750965692521481,
+      "eval_loss": 0.9914345145225525,
+      "eval_mean_token_accuracy": 0.8252089954494091,
+      "eval_num_tokens": 9614034.0,
+      "eval_runtime": 98.5049,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3940
+    },
+    {
+      "entropy": 0.15644967313855887,
+      "epoch": 9.54282267792521,
+      "grad_norm": 0.5718610882759094,
+      "learning_rate": 4.6545198564893463e-07,
+      "loss": 0.08564762473106384,
+      "mean_token_accuracy": 0.9730244860053062,
+      "num_tokens": 9661266.0,
+      "step": 3960
+    },
+    {
+      "epoch": 9.54282267792521,
+      "eval_entropy": 0.27522430820076654,
+      "eval_loss": 0.9908939003944397,
+      "eval_mean_token_accuracy": 0.8251052994406625,
+      "eval_num_tokens": 9661266.0,
+      "eval_runtime": 98.3637,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3960
+    },
+    {
+      "entropy": 0.15196608882397414,
+      "epoch": 9.591073582629674,
+      "grad_norm": 0.752661406993866,
+      "learning_rate": 3.732379748908161e-07,
+      "loss": 0.08072127103805542,
+      "mean_token_accuracy": 0.9749975100159645,
+      "num_tokens": 9710458.0,
+      "step": 3980
+    },
+    {
+      "epoch": 9.591073582629674,
+      "eval_entropy": 0.2750371985890892,
+      "eval_loss": 0.9913897514343262,
+      "eval_mean_token_accuracy": 0.8248539731073915,
+      "eval_num_tokens": 9710458.0,
+      "eval_runtime": 98.4959,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3980
+    },
+    {
+      "entropy": 0.16616767533123494,
+      "epoch": 9.639324487334138,
+      "grad_norm": 0.7107775211334229,
+      "learning_rate": 2.911470833423274e-07,
+      "loss": 0.08988997340202332,
+      "mean_token_accuracy": 0.9714749112725258,
+      "num_tokens": 9755797.0,
+      "step": 4000
+    },
+    {
+      "epoch": 9.639324487334138,
+      "eval_entropy": 0.27523158457172053,
+      "eval_loss": 0.9913759827613831,
+      "eval_mean_token_accuracy": 0.8250464505693885,
+      "eval_num_tokens": 9755797.0,
+      "eval_runtime": 98.5067,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 4000
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.022298921079808e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c41e60387c682fd0581fd1c2048f7be20b429285
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/trainer_state.json
@@ -0,0 +1,4255 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.687575392038601,
+  "eval_steps": 20,
+  "global_step": 4020,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    },
+    {
+      "entropy": 0.17359357613783616,
+      "epoch": 9.012062726176115,
+      "grad_norm": 0.37213996052742004,
+      "learning_rate": 2.138437593000977e-06,
+      "loss": 0.10089871883392335,
+      "mean_token_accuracy": 0.9696638874518566,
+      "num_tokens": 9122597.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.012062726176115,
+      "eval_entropy": 0.2875421148505104,
+      "eval_loss": 0.9357590079307556,
+      "eval_mean_token_accuracy": 0.8264994859025719,
+      "eval_num_tokens": 9122597.0,
+      "eval_runtime": 98.523,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3740
+    },
+    {
+      "entropy": 0.1642639346420765,
+      "epoch": 9.060313630880579,
+      "grad_norm": 0.5511618852615356,
+      "learning_rate": 1.937215494938237e-06,
+      "loss": 0.08524587750434875,
+      "mean_token_accuracy": 0.9754035726189614,
+      "num_tokens": 9171477.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.060313630880579,
+      "eval_entropy": 0.2820046850469675,
+      "eval_loss": 0.9588018655776978,
+      "eval_mean_token_accuracy": 0.8257443647036392,
+      "eval_num_tokens": 9171477.0,
+      "eval_runtime": 98.3992,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1630103051662445,
+      "epoch": 9.108564535585042,
+      "grad_norm": 0.5125291347503662,
+      "learning_rate": 1.7456739289470143e-06,
+      "loss": 0.0895907461643219,
+      "mean_token_accuracy": 0.973100857436657,
+      "num_tokens": 9217543.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.108564535585042,
+      "eval_entropy": 0.2786683516388529,
+      "eval_loss": 0.9743365049362183,
+      "eval_mean_token_accuracy": 0.8253507131940863,
+      "eval_num_tokens": 9217543.0,
+      "eval_runtime": 98.3882,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3780
+    },
+    {
+      "entropy": 0.15873534101992845,
+      "epoch": 9.156815440289506,
+      "grad_norm": 0.5697775483131409,
+      "learning_rate": 1.5638670990247788e-06,
+      "loss": 0.08552997708320617,
+      "mean_token_accuracy": 0.973672679066658,
+      "num_tokens": 9268724.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.156815440289506,
+      "eval_entropy": 0.2781428360704626,
+      "eval_loss": 0.9790021777153015,
+      "eval_mean_token_accuracy": 0.8251898780967413,
+      "eval_num_tokens": 9268724.0,
+      "eval_runtime": 98.5449,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 3800
+    },
+    {
+      "entropy": 0.1606229657307267,
+      "epoch": 9.205066344993968,
+      "grad_norm": 0.5715742707252502,
+      "learning_rate": 1.3918464543538225e-06,
+      "loss": 0.08984670042991638,
+      "mean_token_accuracy": 0.9729854613542557,
+      "num_tokens": 9316805.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.205066344993968,
+      "eval_entropy": 0.2777416408899125,
+      "eval_loss": 0.9808777570724487,
+      "eval_mean_token_accuracy": 0.8252255926641185,
+      "eval_num_tokens": 9316805.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3820
+    },
+    {
+      "entropy": 0.15027269292622805,
+      "epoch": 9.253317249698432,
+      "grad_norm": 0.3359828293323517,
+      "learning_rate": 1.229660674741786e-06,
+      "loss": 0.08229721784591675,
+      "mean_token_accuracy": 0.9754723474383354,
+      "num_tokens": 9369940.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.253317249698432,
+      "eval_entropy": 0.2768587637483404,
+      "eval_loss": 0.9837455749511719,
+      "eval_mean_token_accuracy": 0.825493957889214,
+      "eval_num_tokens": 9369940.0,
+      "eval_runtime": 98.534,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 3840
+    },
+    {
+      "entropy": 0.16767099779099226,
+      "epoch": 9.301568154402895,
+      "grad_norm": 0.5708875060081482,
+      "learning_rate": 1.077355656845822e-06,
+      "loss": 0.08888025283813476,
+      "mean_token_accuracy": 0.9715678468346596,
+      "num_tokens": 9416788.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.301568154402895,
+      "eval_entropy": 0.27612990283229377,
+      "eval_loss": 0.9873999357223511,
+      "eval_mean_token_accuracy": 0.8252144948150335,
+      "eval_num_tokens": 9416788.0,
+      "eval_runtime": 98.3889,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3860
+    },
+    {
+      "entropy": 0.1549418892711401,
+      "epoch": 9.349819059107359,
+      "grad_norm": 0.7601456046104431,
+      "learning_rate": 9.349745011844029e-07,
+      "loss": 0.08411069512367249,
+      "mean_token_accuracy": 0.9741103485226631,
+      "num_tokens": 9467862.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.349819059107359,
+      "eval_entropy": 0.2757877716689967,
+      "eval_loss": 0.9892058968544006,
+      "eval_mean_token_accuracy": 0.8250612017143978,
+      "eval_num_tokens": 9467862.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3880
+    },
+    {
+      "entropy": 0.16111517809331416,
+      "epoch": 9.398069963811821,
+      "grad_norm": 0.6774276494979858,
+      "learning_rate": 8.025574999403695e-07,
+      "loss": 0.08765592575073242,
+      "mean_token_accuracy": 0.9734174907207489,
+      "num_tokens": 9514611.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.398069963811821,
+      "eval_entropy": 0.275664389049739,
+      "eval_loss": 0.989257276058197,
+      "eval_mean_token_accuracy": 0.8252896914321385,
+      "eval_num_tokens": 9514611.0,
+      "eval_runtime": 98.4649,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 3900
+    },
+    {
+      "entropy": 0.15701593235135078,
+      "epoch": 9.446320868516285,
+      "grad_norm": 0.6130298376083374,
+      "learning_rate": 6.801421255587377e-07,
+      "loss": 0.0853091835975647,
+      "mean_token_accuracy": 0.9740088582038879,
+      "num_tokens": 9563437.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.446320868516285,
+      "eval_entropy": 0.27559053881114787,
+      "eval_loss": 0.9901270866394043,
+      "eval_mean_token_accuracy": 0.8251507653948966,
+      "eval_num_tokens": 9563437.0,
+      "eval_runtime": 98.5285,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 3920
+    },
+    {
+      "entropy": 0.15848954878747462,
+      "epoch": 9.494571773220748,
+      "grad_norm": 0.5928868651390076,
+      "learning_rate": 5.677630201424204e-07,
+      "loss": 0.08467000126838684,
+      "mean_token_accuracy": 0.9725634664297104,
+      "num_tokens": 9614034.0,
+      "step": 3940
+    },
+    {
+      "epoch": 9.494571773220748,
+      "eval_entropy": 0.2750965692521481,
+      "eval_loss": 0.9914345145225525,
+      "eval_mean_token_accuracy": 0.8252089954494091,
+      "eval_num_tokens": 9614034.0,
+      "eval_runtime": 98.5049,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3940
+    },
+    {
+      "entropy": 0.15644967313855887,
+      "epoch": 9.54282267792521,
+      "grad_norm": 0.5718610882759094,
+      "learning_rate": 4.6545198564893463e-07,
+      "loss": 0.08564762473106384,
+      "mean_token_accuracy": 0.9730244860053062,
+      "num_tokens": 9661266.0,
+      "step": 3960
+    },
+    {
+      "epoch": 9.54282267792521,
+      "eval_entropy": 0.27522430820076654,
+      "eval_loss": 0.9908939003944397,
+      "eval_mean_token_accuracy": 0.8251052994406625,
+      "eval_num_tokens": 9661266.0,
+      "eval_runtime": 98.3637,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3960
+    },
+    {
+      "entropy": 0.15196608882397414,
+      "epoch": 9.591073582629674,
+      "grad_norm": 0.752661406993866,
+      "learning_rate": 3.732379748908161e-07,
+      "loss": 0.08072127103805542,
+      "mean_token_accuracy": 0.9749975100159645,
+      "num_tokens": 9710458.0,
+      "step": 3980
+    },
+    {
+      "epoch": 9.591073582629674,
+      "eval_entropy": 0.2750371985890892,
+      "eval_loss": 0.9913897514343262,
+      "eval_mean_token_accuracy": 0.8248539731073915,
+      "eval_num_tokens": 9710458.0,
+      "eval_runtime": 98.4959,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3980
+    },
+    {
+      "entropy": 0.16616767533123494,
+      "epoch": 9.639324487334138,
+      "grad_norm": 0.7107775211334229,
+      "learning_rate": 2.911470833423274e-07,
+      "loss": 0.08988997340202332,
+      "mean_token_accuracy": 0.9714749112725258,
+      "num_tokens": 9755797.0,
+      "step": 4000
+    },
+    {
+      "epoch": 9.639324487334138,
+      "eval_entropy": 0.27523158457172053,
+      "eval_loss": 0.9913759827613831,
+      "eval_mean_token_accuracy": 0.8250464505693885,
+      "eval_num_tokens": 9755797.0,
+      "eval_runtime": 98.5067,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 4000
+    },
+    {
+      "entropy": 0.15760798901319503,
+      "epoch": 9.687575392038601,
+      "grad_norm": 0.5563387870788574,
+      "learning_rate": 2.1920254175472834e-07,
+      "loss": 0.0877057135105133,
+      "mean_token_accuracy": 0.9734948605298996,
+      "num_tokens": 9805552.0,
+      "step": 4020
+    },
+    {
+      "epoch": 9.687575392038601,
+      "eval_entropy": 0.27528667726208655,
+      "eval_loss": 0.9914706349372864,
+      "eval_mean_token_accuracy": 0.8250808802883277,
+      "eval_num_tokens": 9805552.0,
+      "eval_runtime": 98.475,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 4020
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.032612367686656e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4040/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4040/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4040/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4040/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4040/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4040/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4040/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4040/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4040/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4040/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4040/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4040/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4040/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4040/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..bce0a0702c20e82c40f173db70cd6b876603ed21
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4040/trainer_state.json
@@ -0,0 +1,4276 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.735826296743063,
+  "eval_steps": 20,
+  "global_step": 4040,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    },
+    {
+      "entropy": 0.17359357613783616,
+      "epoch": 9.012062726176115,
+      "grad_norm": 0.37213996052742004,
+      "learning_rate": 2.138437593000977e-06,
+      "loss": 0.10089871883392335,
+      "mean_token_accuracy": 0.9696638874518566,
+      "num_tokens": 9122597.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.012062726176115,
+      "eval_entropy": 0.2875421148505104,
+      "eval_loss": 0.9357590079307556,
+      "eval_mean_token_accuracy": 0.8264994859025719,
+      "eval_num_tokens": 9122597.0,
+      "eval_runtime": 98.523,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3740
+    },
+    {
+      "entropy": 0.1642639346420765,
+      "epoch": 9.060313630880579,
+      "grad_norm": 0.5511618852615356,
+      "learning_rate": 1.937215494938237e-06,
+      "loss": 0.08524587750434875,
+      "mean_token_accuracy": 0.9754035726189614,
+      "num_tokens": 9171477.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.060313630880579,
+      "eval_entropy": 0.2820046850469675,
+      "eval_loss": 0.9588018655776978,
+      "eval_mean_token_accuracy": 0.8257443647036392,
+      "eval_num_tokens": 9171477.0,
+      "eval_runtime": 98.3992,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1630103051662445,
+      "epoch": 9.108564535585042,
+      "grad_norm": 0.5125291347503662,
+      "learning_rate": 1.7456739289470143e-06,
+      "loss": 0.0895907461643219,
+      "mean_token_accuracy": 0.973100857436657,
+      "num_tokens": 9217543.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.108564535585042,
+      "eval_entropy": 0.2786683516388529,
+      "eval_loss": 0.9743365049362183,
+      "eval_mean_token_accuracy": 0.8253507131940863,
+      "eval_num_tokens": 9217543.0,
+      "eval_runtime": 98.3882,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3780
+    },
+    {
+      "entropy": 0.15873534101992845,
+      "epoch": 9.156815440289506,
+      "grad_norm": 0.5697775483131409,
+      "learning_rate": 1.5638670990247788e-06,
+      "loss": 0.08552997708320617,
+      "mean_token_accuracy": 0.973672679066658,
+      "num_tokens": 9268724.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.156815440289506,
+      "eval_entropy": 0.2781428360704626,
+      "eval_loss": 0.9790021777153015,
+      "eval_mean_token_accuracy": 0.8251898780967413,
+      "eval_num_tokens": 9268724.0,
+      "eval_runtime": 98.5449,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 3800
+    },
+    {
+      "entropy": 0.1606229657307267,
+      "epoch": 9.205066344993968,
+      "grad_norm": 0.5715742707252502,
+      "learning_rate": 1.3918464543538225e-06,
+      "loss": 0.08984670042991638,
+      "mean_token_accuracy": 0.9729854613542557,
+      "num_tokens": 9316805.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.205066344993968,
+      "eval_entropy": 0.2777416408899125,
+      "eval_loss": 0.9808777570724487,
+      "eval_mean_token_accuracy": 0.8252255926641185,
+      "eval_num_tokens": 9316805.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3820
+    },
+    {
+      "entropy": 0.15027269292622805,
+      "epoch": 9.253317249698432,
+      "grad_norm": 0.3359828293323517,
+      "learning_rate": 1.229660674741786e-06,
+      "loss": 0.08229721784591675,
+      "mean_token_accuracy": 0.9754723474383354,
+      "num_tokens": 9369940.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.253317249698432,
+      "eval_entropy": 0.2768587637483404,
+      "eval_loss": 0.9837455749511719,
+      "eval_mean_token_accuracy": 0.825493957889214,
+      "eval_num_tokens": 9369940.0,
+      "eval_runtime": 98.534,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 3840
+    },
+    {
+      "entropy": 0.16767099779099226,
+      "epoch": 9.301568154402895,
+      "grad_norm": 0.5708875060081482,
+      "learning_rate": 1.077355656845822e-06,
+      "loss": 0.08888025283813476,
+      "mean_token_accuracy": 0.9715678468346596,
+      "num_tokens": 9416788.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.301568154402895,
+      "eval_entropy": 0.27612990283229377,
+      "eval_loss": 0.9873999357223511,
+      "eval_mean_token_accuracy": 0.8252144948150335,
+      "eval_num_tokens": 9416788.0,
+      "eval_runtime": 98.3889,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3860
+    },
+    {
+      "entropy": 0.1549418892711401,
+      "epoch": 9.349819059107359,
+      "grad_norm": 0.7601456046104431,
+      "learning_rate": 9.349745011844029e-07,
+      "loss": 0.08411069512367249,
+      "mean_token_accuracy": 0.9741103485226631,
+      "num_tokens": 9467862.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.349819059107359,
+      "eval_entropy": 0.2757877716689967,
+      "eval_loss": 0.9892058968544006,
+      "eval_mean_token_accuracy": 0.8250612017143978,
+      "eval_num_tokens": 9467862.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3880
+    },
+    {
+      "entropy": 0.16111517809331416,
+      "epoch": 9.398069963811821,
+      "grad_norm": 0.6774276494979858,
+      "learning_rate": 8.025574999403695e-07,
+      "loss": 0.08765592575073242,
+      "mean_token_accuracy": 0.9734174907207489,
+      "num_tokens": 9514611.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.398069963811821,
+      "eval_entropy": 0.275664389049739,
+      "eval_loss": 0.989257276058197,
+      "eval_mean_token_accuracy": 0.8252896914321385,
+      "eval_num_tokens": 9514611.0,
+      "eval_runtime": 98.4649,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 3900
+    },
+    {
+      "entropy": 0.15701593235135078,
+      "epoch": 9.446320868516285,
+      "grad_norm": 0.6130298376083374,
+      "learning_rate": 6.801421255587377e-07,
+      "loss": 0.0853091835975647,
+      "mean_token_accuracy": 0.9740088582038879,
+      "num_tokens": 9563437.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.446320868516285,
+      "eval_entropy": 0.27559053881114787,
+      "eval_loss": 0.9901270866394043,
+      "eval_mean_token_accuracy": 0.8251507653948966,
+      "eval_num_tokens": 9563437.0,
+      "eval_runtime": 98.5285,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 3920
+    },
+    {
+      "entropy": 0.15848954878747462,
+      "epoch": 9.494571773220748,
+      "grad_norm": 0.5928868651390076,
+      "learning_rate": 5.677630201424204e-07,
+      "loss": 0.08467000126838684,
+      "mean_token_accuracy": 0.9725634664297104,
+      "num_tokens": 9614034.0,
+      "step": 3940
+    },
+    {
+      "epoch": 9.494571773220748,
+      "eval_entropy": 0.2750965692521481,
+      "eval_loss": 0.9914345145225525,
+      "eval_mean_token_accuracy": 0.8252089954494091,
+      "eval_num_tokens": 9614034.0,
+      "eval_runtime": 98.5049,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3940
+    },
+    {
+      "entropy": 0.15644967313855887,
+      "epoch": 9.54282267792521,
+      "grad_norm": 0.5718610882759094,
+      "learning_rate": 4.6545198564893463e-07,
+      "loss": 0.08564762473106384,
+      "mean_token_accuracy": 0.9730244860053062,
+      "num_tokens": 9661266.0,
+      "step": 3960
+    },
+    {
+      "epoch": 9.54282267792521,
+      "eval_entropy": 0.27522430820076654,
+      "eval_loss": 0.9908939003944397,
+      "eval_mean_token_accuracy": 0.8251052994406625,
+      "eval_num_tokens": 9661266.0,
+      "eval_runtime": 98.3637,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3960
+    },
+    {
+      "entropy": 0.15196608882397414,
+      "epoch": 9.591073582629674,
+      "grad_norm": 0.752661406993866,
+      "learning_rate": 3.732379748908161e-07,
+      "loss": 0.08072127103805542,
+      "mean_token_accuracy": 0.9749975100159645,
+      "num_tokens": 9710458.0,
+      "step": 3980
+    },
+    {
+      "epoch": 9.591073582629674,
+      "eval_entropy": 0.2750371985890892,
+      "eval_loss": 0.9913897514343262,
+      "eval_mean_token_accuracy": 0.8248539731073915,
+      "eval_num_tokens": 9710458.0,
+      "eval_runtime": 98.4959,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3980
+    },
+    {
+      "entropy": 0.16616767533123494,
+      "epoch": 9.639324487334138,
+      "grad_norm": 0.7107775211334229,
+      "learning_rate": 2.911470833423274e-07,
+      "loss": 0.08988997340202332,
+      "mean_token_accuracy": 0.9714749112725258,
+      "num_tokens": 9755797.0,
+      "step": 4000
+    },
+    {
+      "epoch": 9.639324487334138,
+      "eval_entropy": 0.27523158457172053,
+      "eval_loss": 0.9913759827613831,
+      "eval_mean_token_accuracy": 0.8250464505693885,
+      "eval_num_tokens": 9755797.0,
+      "eval_runtime": 98.5067,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 4000
+    },
+    {
+      "entropy": 0.15760798901319503,
+      "epoch": 9.687575392038601,
+      "grad_norm": 0.5563387870788574,
+      "learning_rate": 2.1920254175472834e-07,
+      "loss": 0.0877057135105133,
+      "mean_token_accuracy": 0.9734948605298996,
+      "num_tokens": 9805552.0,
+      "step": 4020
+    },
+    {
+      "epoch": 9.687575392038601,
+      "eval_entropy": 0.27528667726208655,
+      "eval_loss": 0.9914706349372864,
+      "eval_mean_token_accuracy": 0.8250808802883277,
+      "eval_num_tokens": 9805552.0,
+      "eval_runtime": 98.475,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 4020
+    },
+    {
+      "entropy": 0.161653464846313,
+      "epoch": 9.735826296743063,
+      "grad_norm": 0.5896415114402771,
+      "learning_rate": 1.5742470958227596e-07,
+      "loss": 0.08775997757911683,
+      "mean_token_accuracy": 0.9718310445547104,
+      "num_tokens": 9852843.0,
+      "step": 4040
+    },
+    {
+      "epoch": 9.735826296743063,
+      "eval_entropy": 0.27531274182073184,
+      "eval_loss": 0.9915319085121155,
+      "eval_mean_token_accuracy": 0.825061980258213,
+      "eval_num_tokens": 9852843.0,
+      "eval_runtime": 98.3873,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 4040
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.042777414458368e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4060/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4060/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4060/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4060/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4060/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4060/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4060/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4060/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4060/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4060/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4060/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4060/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4060/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4060/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..bf2ba2c14f56c7b74070f9c21bca365e38f36b49
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4060/trainer_state.json
@@ -0,0 +1,4297 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.784077201447527,
+  "eval_steps": 20,
+  "global_step": 4060,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    },
+    {
+      "entropy": 0.17359357613783616,
+      "epoch": 9.012062726176115,
+      "grad_norm": 0.37213996052742004,
+      "learning_rate": 2.138437593000977e-06,
+      "loss": 0.10089871883392335,
+      "mean_token_accuracy": 0.9696638874518566,
+      "num_tokens": 9122597.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.012062726176115,
+      "eval_entropy": 0.2875421148505104,
+      "eval_loss": 0.9357590079307556,
+      "eval_mean_token_accuracy": 0.8264994859025719,
+      "eval_num_tokens": 9122597.0,
+      "eval_runtime": 98.523,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3740
+    },
+    {
+      "entropy": 0.1642639346420765,
+      "epoch": 9.060313630880579,
+      "grad_norm": 0.5511618852615356,
+      "learning_rate": 1.937215494938237e-06,
+      "loss": 0.08524587750434875,
+      "mean_token_accuracy": 0.9754035726189614,
+      "num_tokens": 9171477.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.060313630880579,
+      "eval_entropy": 0.2820046850469675,
+      "eval_loss": 0.9588018655776978,
+      "eval_mean_token_accuracy": 0.8257443647036392,
+      "eval_num_tokens": 9171477.0,
+      "eval_runtime": 98.3992,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1630103051662445,
+      "epoch": 9.108564535585042,
+      "grad_norm": 0.5125291347503662,
+      "learning_rate": 1.7456739289470143e-06,
+      "loss": 0.0895907461643219,
+      "mean_token_accuracy": 0.973100857436657,
+      "num_tokens": 9217543.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.108564535585042,
+      "eval_entropy": 0.2786683516388529,
+      "eval_loss": 0.9743365049362183,
+      "eval_mean_token_accuracy": 0.8253507131940863,
+      "eval_num_tokens": 9217543.0,
+      "eval_runtime": 98.3882,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3780
+    },
+    {
+      "entropy": 0.15873534101992845,
+      "epoch": 9.156815440289506,
+      "grad_norm": 0.5697775483131409,
+      "learning_rate": 1.5638670990247788e-06,
+      "loss": 0.08552997708320617,
+      "mean_token_accuracy": 0.973672679066658,
+      "num_tokens": 9268724.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.156815440289506,
+      "eval_entropy": 0.2781428360704626,
+      "eval_loss": 0.9790021777153015,
+      "eval_mean_token_accuracy": 0.8251898780967413,
+      "eval_num_tokens": 9268724.0,
+      "eval_runtime": 98.5449,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 3800
+    },
+    {
+      "entropy": 0.1606229657307267,
+      "epoch": 9.205066344993968,
+      "grad_norm": 0.5715742707252502,
+      "learning_rate": 1.3918464543538225e-06,
+      "loss": 0.08984670042991638,
+      "mean_token_accuracy": 0.9729854613542557,
+      "num_tokens": 9316805.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.205066344993968,
+      "eval_entropy": 0.2777416408899125,
+      "eval_loss": 0.9808777570724487,
+      "eval_mean_token_accuracy": 0.8252255926641185,
+      "eval_num_tokens": 9316805.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3820
+    },
+    {
+      "entropy": 0.15027269292622805,
+      "epoch": 9.253317249698432,
+      "grad_norm": 0.3359828293323517,
+      "learning_rate": 1.229660674741786e-06,
+      "loss": 0.08229721784591675,
+      "mean_token_accuracy": 0.9754723474383354,
+      "num_tokens": 9369940.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.253317249698432,
+      "eval_entropy": 0.2768587637483404,
+      "eval_loss": 0.9837455749511719,
+      "eval_mean_token_accuracy": 0.825493957889214,
+      "eval_num_tokens": 9369940.0,
+      "eval_runtime": 98.534,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 3840
+    },
+    {
+      "entropy": 0.16767099779099226,
+      "epoch": 9.301568154402895,
+      "grad_norm": 0.5708875060081482,
+      "learning_rate": 1.077355656845822e-06,
+      "loss": 0.08888025283813476,
+      "mean_token_accuracy": 0.9715678468346596,
+      "num_tokens": 9416788.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.301568154402895,
+      "eval_entropy": 0.27612990283229377,
+      "eval_loss": 0.9873999357223511,
+      "eval_mean_token_accuracy": 0.8252144948150335,
+      "eval_num_tokens": 9416788.0,
+      "eval_runtime": 98.3889,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3860
+    },
+    {
+      "entropy": 0.1549418892711401,
+      "epoch": 9.349819059107359,
+      "grad_norm": 0.7601456046104431,
+      "learning_rate": 9.349745011844029e-07,
+      "loss": 0.08411069512367249,
+      "mean_token_accuracy": 0.9741103485226631,
+      "num_tokens": 9467862.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.349819059107359,
+      "eval_entropy": 0.2757877716689967,
+      "eval_loss": 0.9892058968544006,
+      "eval_mean_token_accuracy": 0.8250612017143978,
+      "eval_num_tokens": 9467862.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3880
+    },
+    {
+      "entropy": 0.16111517809331416,
+      "epoch": 9.398069963811821,
+      "grad_norm": 0.6774276494979858,
+      "learning_rate": 8.025574999403695e-07,
+      "loss": 0.08765592575073242,
+      "mean_token_accuracy": 0.9734174907207489,
+      "num_tokens": 9514611.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.398069963811821,
+      "eval_entropy": 0.275664389049739,
+      "eval_loss": 0.989257276058197,
+      "eval_mean_token_accuracy": 0.8252896914321385,
+      "eval_num_tokens": 9514611.0,
+      "eval_runtime": 98.4649,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 3900
+    },
+    {
+      "entropy": 0.15701593235135078,
+      "epoch": 9.446320868516285,
+      "grad_norm": 0.6130298376083374,
+      "learning_rate": 6.801421255587377e-07,
+      "loss": 0.0853091835975647,
+      "mean_token_accuracy": 0.9740088582038879,
+      "num_tokens": 9563437.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.446320868516285,
+      "eval_entropy": 0.27559053881114787,
+      "eval_loss": 0.9901270866394043,
+      "eval_mean_token_accuracy": 0.8251507653948966,
+      "eval_num_tokens": 9563437.0,
+      "eval_runtime": 98.5285,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 3920
+    },
+    {
+      "entropy": 0.15848954878747462,
+      "epoch": 9.494571773220748,
+      "grad_norm": 0.5928868651390076,
+      "learning_rate": 5.677630201424204e-07,
+      "loss": 0.08467000126838684,
+      "mean_token_accuracy": 0.9725634664297104,
+      "num_tokens": 9614034.0,
+      "step": 3940
+    },
+    {
+      "epoch": 9.494571773220748,
+      "eval_entropy": 0.2750965692521481,
+      "eval_loss": 0.9914345145225525,
+      "eval_mean_token_accuracy": 0.8252089954494091,
+      "eval_num_tokens": 9614034.0,
+      "eval_runtime": 98.5049,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3940
+    },
+    {
+      "entropy": 0.15644967313855887,
+      "epoch": 9.54282267792521,
+      "grad_norm": 0.5718610882759094,
+      "learning_rate": 4.6545198564893463e-07,
+      "loss": 0.08564762473106384,
+      "mean_token_accuracy": 0.9730244860053062,
+      "num_tokens": 9661266.0,
+      "step": 3960
+    },
+    {
+      "epoch": 9.54282267792521,
+      "eval_entropy": 0.27522430820076654,
+      "eval_loss": 0.9908939003944397,
+      "eval_mean_token_accuracy": 0.8251052994406625,
+      "eval_num_tokens": 9661266.0,
+      "eval_runtime": 98.3637,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3960
+    },
+    {
+      "entropy": 0.15196608882397414,
+      "epoch": 9.591073582629674,
+      "grad_norm": 0.752661406993866,
+      "learning_rate": 3.732379748908161e-07,
+      "loss": 0.08072127103805542,
+      "mean_token_accuracy": 0.9749975100159645,
+      "num_tokens": 9710458.0,
+      "step": 3980
+    },
+    {
+      "epoch": 9.591073582629674,
+      "eval_entropy": 0.2750371985890892,
+      "eval_loss": 0.9913897514343262,
+      "eval_mean_token_accuracy": 0.8248539731073915,
+      "eval_num_tokens": 9710458.0,
+      "eval_runtime": 98.4959,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3980
+    },
+    {
+      "entropy": 0.16616767533123494,
+      "epoch": 9.639324487334138,
+      "grad_norm": 0.7107775211334229,
+      "learning_rate": 2.911470833423274e-07,
+      "loss": 0.08988997340202332,
+      "mean_token_accuracy": 0.9714749112725258,
+      "num_tokens": 9755797.0,
+      "step": 4000
+    },
+    {
+      "epoch": 9.639324487334138,
+      "eval_entropy": 0.27523158457172053,
+      "eval_loss": 0.9913759827613831,
+      "eval_mean_token_accuracy": 0.8250464505693885,
+      "eval_num_tokens": 9755797.0,
+      "eval_runtime": 98.5067,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 4000
+    },
+    {
+      "entropy": 0.15760798901319503,
+      "epoch": 9.687575392038601,
+      "grad_norm": 0.5563387870788574,
+      "learning_rate": 2.1920254175472834e-07,
+      "loss": 0.0877057135105133,
+      "mean_token_accuracy": 0.9734948605298996,
+      "num_tokens": 9805552.0,
+      "step": 4020
+    },
+    {
+      "epoch": 9.687575392038601,
+      "eval_entropy": 0.27528667726208655,
+      "eval_loss": 0.9914706349372864,
+      "eval_mean_token_accuracy": 0.8250808802883277,
+      "eval_num_tokens": 9805552.0,
+      "eval_runtime": 98.475,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 4020
+    },
+    {
+      "entropy": 0.161653464846313,
+      "epoch": 9.735826296743063,
+      "grad_norm": 0.5896415114402771,
+      "learning_rate": 1.5742470958227596e-07,
+      "loss": 0.08775997757911683,
+      "mean_token_accuracy": 0.9718310445547104,
+      "num_tokens": 9852843.0,
+      "step": 4040
+    },
+    {
+      "epoch": 9.735826296743063,
+      "eval_entropy": 0.27531274182073184,
+      "eval_loss": 0.9915319085121155,
+      "eval_mean_token_accuracy": 0.825061980258213,
+      "eval_num_tokens": 9852843.0,
+      "eval_runtime": 98.3873,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 4040
+    },
+    {
+      "entropy": 0.16010624952614308,
+      "epoch": 9.784077201447527,
+      "grad_norm": 0.6022347211837769,
+      "learning_rate": 1.058310692207063e-07,
+      "loss": 0.08761411905288696,
+      "mean_token_accuracy": 0.9729624018073082,
+      "num_tokens": 9900151.0,
+      "step": 4060
+    },
+    {
+      "epoch": 9.784077201447527,
+      "eval_entropy": 0.27524732446737504,
+      "eval_loss": 0.991264820098877,
+      "eval_mean_token_accuracy": 0.8251047007153544,
+      "eval_num_tokens": 9900151.0,
+      "eval_runtime": 98.4834,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 4060
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.052873390339072e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4080/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4080/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4080/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4080/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4080/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4080/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4080/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4080/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4080/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4080/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4080/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4080/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4080/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4080/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..3662f3e43de26b7dbc0e70d67dd91217356be85d
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4080/trainer_state.json
@@ -0,0 +1,4318 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.83232810615199,
+  "eval_steps": 20,
+  "global_step": 4080,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    },
+    {
+      "entropy": 0.17359357613783616,
+      "epoch": 9.012062726176115,
+      "grad_norm": 0.37213996052742004,
+      "learning_rate": 2.138437593000977e-06,
+      "loss": 0.10089871883392335,
+      "mean_token_accuracy": 0.9696638874518566,
+      "num_tokens": 9122597.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.012062726176115,
+      "eval_entropy": 0.2875421148505104,
+      "eval_loss": 0.9357590079307556,
+      "eval_mean_token_accuracy": 0.8264994859025719,
+      "eval_num_tokens": 9122597.0,
+      "eval_runtime": 98.523,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3740
+    },
+    {
+      "entropy": 0.1642639346420765,
+      "epoch": 9.060313630880579,
+      "grad_norm": 0.5511618852615356,
+      "learning_rate": 1.937215494938237e-06,
+      "loss": 0.08524587750434875,
+      "mean_token_accuracy": 0.9754035726189614,
+      "num_tokens": 9171477.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.060313630880579,
+      "eval_entropy": 0.2820046850469675,
+      "eval_loss": 0.9588018655776978,
+      "eval_mean_token_accuracy": 0.8257443647036392,
+      "eval_num_tokens": 9171477.0,
+      "eval_runtime": 98.3992,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1630103051662445,
+      "epoch": 9.108564535585042,
+      "grad_norm": 0.5125291347503662,
+      "learning_rate": 1.7456739289470143e-06,
+      "loss": 0.0895907461643219,
+      "mean_token_accuracy": 0.973100857436657,
+      "num_tokens": 9217543.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.108564535585042,
+      "eval_entropy": 0.2786683516388529,
+      "eval_loss": 0.9743365049362183,
+      "eval_mean_token_accuracy": 0.8253507131940863,
+      "eval_num_tokens": 9217543.0,
+      "eval_runtime": 98.3882,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3780
+    },
+    {
+      "entropy": 0.15873534101992845,
+      "epoch": 9.156815440289506,
+      "grad_norm": 0.5697775483131409,
+      "learning_rate": 1.5638670990247788e-06,
+      "loss": 0.08552997708320617,
+      "mean_token_accuracy": 0.973672679066658,
+      "num_tokens": 9268724.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.156815440289506,
+      "eval_entropy": 0.2781428360704626,
+      "eval_loss": 0.9790021777153015,
+      "eval_mean_token_accuracy": 0.8251898780967413,
+      "eval_num_tokens": 9268724.0,
+      "eval_runtime": 98.5449,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 3800
+    },
+    {
+      "entropy": 0.1606229657307267,
+      "epoch": 9.205066344993968,
+      "grad_norm": 0.5715742707252502,
+      "learning_rate": 1.3918464543538225e-06,
+      "loss": 0.08984670042991638,
+      "mean_token_accuracy": 0.9729854613542557,
+      "num_tokens": 9316805.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.205066344993968,
+      "eval_entropy": 0.2777416408899125,
+      "eval_loss": 0.9808777570724487,
+      "eval_mean_token_accuracy": 0.8252255926641185,
+      "eval_num_tokens": 9316805.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3820
+    },
+    {
+      "entropy": 0.15027269292622805,
+      "epoch": 9.253317249698432,
+      "grad_norm": 0.3359828293323517,
+      "learning_rate": 1.229660674741786e-06,
+      "loss": 0.08229721784591675,
+      "mean_token_accuracy": 0.9754723474383354,
+      "num_tokens": 9369940.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.253317249698432,
+      "eval_entropy": 0.2768587637483404,
+      "eval_loss": 0.9837455749511719,
+      "eval_mean_token_accuracy": 0.825493957889214,
+      "eval_num_tokens": 9369940.0,
+      "eval_runtime": 98.534,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 3840
+    },
+    {
+      "entropy": 0.16767099779099226,
+      "epoch": 9.301568154402895,
+      "grad_norm": 0.5708875060081482,
+      "learning_rate": 1.077355656845822e-06,
+      "loss": 0.08888025283813476,
+      "mean_token_accuracy": 0.9715678468346596,
+      "num_tokens": 9416788.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.301568154402895,
+      "eval_entropy": 0.27612990283229377,
+      "eval_loss": 0.9873999357223511,
+      "eval_mean_token_accuracy": 0.8252144948150335,
+      "eval_num_tokens": 9416788.0,
+      "eval_runtime": 98.3889,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3860
+    },
+    {
+      "entropy": 0.1549418892711401,
+      "epoch": 9.349819059107359,
+      "grad_norm": 0.7601456046104431,
+      "learning_rate": 9.349745011844029e-07,
+      "loss": 0.08411069512367249,
+      "mean_token_accuracy": 0.9741103485226631,
+      "num_tokens": 9467862.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.349819059107359,
+      "eval_entropy": 0.2757877716689967,
+      "eval_loss": 0.9892058968544006,
+      "eval_mean_token_accuracy": 0.8250612017143978,
+      "eval_num_tokens": 9467862.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3880
+    },
+    {
+      "entropy": 0.16111517809331416,
+      "epoch": 9.398069963811821,
+      "grad_norm": 0.6774276494979858,
+      "learning_rate": 8.025574999403695e-07,
+      "loss": 0.08765592575073242,
+      "mean_token_accuracy": 0.9734174907207489,
+      "num_tokens": 9514611.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.398069963811821,
+      "eval_entropy": 0.275664389049739,
+      "eval_loss": 0.989257276058197,
+      "eval_mean_token_accuracy": 0.8252896914321385,
+      "eval_num_tokens": 9514611.0,
+      "eval_runtime": 98.4649,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 3900
+    },
+    {
+      "entropy": 0.15701593235135078,
+      "epoch": 9.446320868516285,
+      "grad_norm": 0.6130298376083374,
+      "learning_rate": 6.801421255587377e-07,
+      "loss": 0.0853091835975647,
+      "mean_token_accuracy": 0.9740088582038879,
+      "num_tokens": 9563437.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.446320868516285,
+      "eval_entropy": 0.27559053881114787,
+      "eval_loss": 0.9901270866394043,
+      "eval_mean_token_accuracy": 0.8251507653948966,
+      "eval_num_tokens": 9563437.0,
+      "eval_runtime": 98.5285,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 3920
+    },
+    {
+      "entropy": 0.15848954878747462,
+      "epoch": 9.494571773220748,
+      "grad_norm": 0.5928868651390076,
+      "learning_rate": 5.677630201424204e-07,
+      "loss": 0.08467000126838684,
+      "mean_token_accuracy": 0.9725634664297104,
+      "num_tokens": 9614034.0,
+      "step": 3940
+    },
+    {
+      "epoch": 9.494571773220748,
+      "eval_entropy": 0.2750965692521481,
+      "eval_loss": 0.9914345145225525,
+      "eval_mean_token_accuracy": 0.8252089954494091,
+      "eval_num_tokens": 9614034.0,
+      "eval_runtime": 98.5049,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3940
+    },
+    {
+      "entropy": 0.15644967313855887,
+      "epoch": 9.54282267792521,
+      "grad_norm": 0.5718610882759094,
+      "learning_rate": 4.6545198564893463e-07,
+      "loss": 0.08564762473106384,
+      "mean_token_accuracy": 0.9730244860053062,
+      "num_tokens": 9661266.0,
+      "step": 3960
+    },
+    {
+      "epoch": 9.54282267792521,
+      "eval_entropy": 0.27522430820076654,
+      "eval_loss": 0.9908939003944397,
+      "eval_mean_token_accuracy": 0.8251052994406625,
+      "eval_num_tokens": 9661266.0,
+      "eval_runtime": 98.3637,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3960
+    },
+    {
+      "entropy": 0.15196608882397414,
+      "epoch": 9.591073582629674,
+      "grad_norm": 0.752661406993866,
+      "learning_rate": 3.732379748908161e-07,
+      "loss": 0.08072127103805542,
+      "mean_token_accuracy": 0.9749975100159645,
+      "num_tokens": 9710458.0,
+      "step": 3980
+    },
+    {
+      "epoch": 9.591073582629674,
+      "eval_entropy": 0.2750371985890892,
+      "eval_loss": 0.9913897514343262,
+      "eval_mean_token_accuracy": 0.8248539731073915,
+      "eval_num_tokens": 9710458.0,
+      "eval_runtime": 98.4959,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3980
+    },
+    {
+      "entropy": 0.16616767533123494,
+      "epoch": 9.639324487334138,
+      "grad_norm": 0.7107775211334229,
+      "learning_rate": 2.911470833423274e-07,
+      "loss": 0.08988997340202332,
+      "mean_token_accuracy": 0.9714749112725258,
+      "num_tokens": 9755797.0,
+      "step": 4000
+    },
+    {
+      "epoch": 9.639324487334138,
+      "eval_entropy": 0.27523158457172053,
+      "eval_loss": 0.9913759827613831,
+      "eval_mean_token_accuracy": 0.8250464505693885,
+      "eval_num_tokens": 9755797.0,
+      "eval_runtime": 98.5067,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 4000
+    },
+    {
+      "entropy": 0.15760798901319503,
+      "epoch": 9.687575392038601,
+      "grad_norm": 0.5563387870788574,
+      "learning_rate": 2.1920254175472834e-07,
+      "loss": 0.0877057135105133,
+      "mean_token_accuracy": 0.9734948605298996,
+      "num_tokens": 9805552.0,
+      "step": 4020
+    },
+    {
+      "epoch": 9.687575392038601,
+      "eval_entropy": 0.27528667726208655,
+      "eval_loss": 0.9914706349372864,
+      "eval_mean_token_accuracy": 0.8250808802883277,
+      "eval_num_tokens": 9805552.0,
+      "eval_runtime": 98.475,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 4020
+    },
+    {
+      "entropy": 0.161653464846313,
+      "epoch": 9.735826296743063,
+      "grad_norm": 0.5896415114402771,
+      "learning_rate": 1.5742470958227596e-07,
+      "loss": 0.08775997757911683,
+      "mean_token_accuracy": 0.9718310445547104,
+      "num_tokens": 9852843.0,
+      "step": 4040
+    },
+    {
+      "epoch": 9.735826296743063,
+      "eval_entropy": 0.27531274182073184,
+      "eval_loss": 0.9915319085121155,
+      "eval_mean_token_accuracy": 0.825061980258213,
+      "eval_num_tokens": 9852843.0,
+      "eval_runtime": 98.3873,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 4040
+    },
+    {
+      "entropy": 0.16010624952614308,
+      "epoch": 9.784077201447527,
+      "grad_norm": 0.6022347211837769,
+      "learning_rate": 1.058310692207063e-07,
+      "loss": 0.08761411905288696,
+      "mean_token_accuracy": 0.9729624018073082,
+      "num_tokens": 9900151.0,
+      "step": 4060
+    },
+    {
+      "epoch": 9.784077201447527,
+      "eval_entropy": 0.27524732446737504,
+      "eval_loss": 0.991264820098877,
+      "eval_mean_token_accuracy": 0.8251047007153544,
+      "eval_num_tokens": 9900151.0,
+      "eval_runtime": 98.4834,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 4060
+    },
+    {
+      "entropy": 0.159021489135921,
+      "epoch": 9.83232810615199,
+      "grad_norm": 0.48451653122901917,
+      "learning_rate": 6.443622105994901e-08,
+      "loss": 0.08291574716567993,
+      "mean_token_accuracy": 0.9726693674921989,
+      "num_tokens": 9952926.0,
+      "step": 4080
+    },
+    {
+      "epoch": 9.83232810615199,
+      "eval_entropy": 0.27521907144717955,
+      "eval_loss": 0.9918096661567688,
+      "eval_mean_token_accuracy": 0.8250070894032382,
+      "eval_num_tokens": 9952926.0,
+      "eval_runtime": 98.3568,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 4080
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.064060823071744e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4100/README.md b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4100/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4100/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4100/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4100/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccfc07021d124c37907d00e2fac38d632c0b509
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4100/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.009940685857937281,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4100/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4100/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4100/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4100/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4100/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4100/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4100/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4100/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b92e3021306f27a5597414040314001c25ef33c9
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4100/trainer_state.json
@@ -0,0 +1,4339 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.880579010856454,
+  "eval_steps": 20,
+  "global_step": 4100,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8880569130182265,
+      "epoch": 0.04825090470446321,
+      "grad_norm": 0.5548511743545532,
+      "learning_rate": 3.3097059104127273e-06,
+      "loss": 1.863978958129883,
+      "mean_token_accuracy": 0.621908786892891,
+      "num_tokens": 50237.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04825090470446321,
+      "eval_entropy": 1.8981187470843284,
+      "eval_loss": 1.9387791156768799,
+      "eval_mean_token_accuracy": 0.6191755044326354,
+      "eval_num_tokens": 50237.0,
+      "eval_runtime": 99.1278,
+      "eval_samples_per_second": 14.325,
+      "eval_steps_per_second": 1.796,
+      "step": 20
+    },
+    {
+      "entropy": 1.845469006896019,
+      "epoch": 0.09650180940892641,
+      "grad_norm": 0.3859543204307556,
+      "learning_rate": 6.793606868741914e-06,
+      "loss": 1.8429210662841797,
+      "mean_token_accuracy": 0.6286612510681152,
+      "num_tokens": 101098.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09650180940892641,
+      "eval_entropy": 1.8266146584842982,
+      "eval_loss": 1.8124289512634277,
+      "eval_mean_token_accuracy": 0.636085176400924,
+      "eval_num_tokens": 101098.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 40
+    },
+    {
+      "entropy": 1.6761731654405594,
+      "epoch": 0.14475271411338964,
+      "grad_norm": 0.763229489326477,
+      "learning_rate": 1.0277507827071102e-05,
+      "loss": 1.59182710647583,
+      "mean_token_accuracy": 0.6615026861429214,
+      "num_tokens": 150385.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14475271411338964,
+      "eval_entropy": 1.448009472884489,
+      "eval_loss": 1.4045010805130005,
+      "eval_mean_token_accuracy": 0.6956760652279586,
+      "eval_num_tokens": 150385.0,
+      "eval_runtime": 98.6634,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 60
+    },
+    {
+      "entropy": 1.2918378412723541,
+      "epoch": 0.19300361881785283,
+      "grad_norm": 0.6131361126899719,
+      "learning_rate": 1.3761408785400288e-05,
+      "loss": 1.2077620506286622,
+      "mean_token_accuracy": 0.715845774114132,
+      "num_tokens": 197354.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19300361881785283,
+      "eval_entropy": 1.0959123356288738,
+      "eval_loss": 1.025644063949585,
+      "eval_mean_token_accuracy": 0.7453202035989654,
+      "eval_num_tokens": 197354.0,
+      "eval_runtime": 98.6386,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 80
+    },
+    {
+      "entropy": 0.9881208389997482,
+      "epoch": 0.24125452352231605,
+      "grad_norm": 0.4983534514904022,
+      "learning_rate": 1.7245309743729473e-05,
+      "loss": 0.9109245300292969,
+      "mean_token_accuracy": 0.7622973442077636,
+      "num_tokens": 248557.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24125452352231605,
+      "eval_entropy": 0.9157560136211053,
+      "eval_loss": 0.8661546111106873,
+      "eval_mean_token_accuracy": 0.7745042727904373,
+      "eval_num_tokens": 248557.0,
+      "eval_runtime": 98.5282,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 100
+    },
+    {
+      "entropy": 0.8518766984343529,
+      "epoch": 0.28950542822677927,
+      "grad_norm": 0.635079026222229,
+      "learning_rate": 2.0729210702058663e-05,
+      "loss": 0.8058656692504883,
+      "mean_token_accuracy": 0.7847469851374627,
+      "num_tokens": 298041.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28950542822677927,
+      "eval_entropy": 0.838603059227547,
+      "eval_loss": 0.7992516756057739,
+      "eval_mean_token_accuracy": 0.7849764954508021,
+      "eval_num_tokens": 298041.0,
+      "eval_runtime": 98.5627,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 120
+    },
+    {
+      "entropy": 0.8069365546107292,
+      "epoch": 0.33775633293124246,
+      "grad_norm": 0.5280472636222839,
+      "learning_rate": 2.421311166038785e-05,
+      "loss": 0.7562496662139893,
+      "mean_token_accuracy": 0.7897997036576271,
+      "num_tokens": 348955.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33775633293124246,
+      "eval_entropy": 0.8161682168419442,
+      "eval_loss": 0.7599642276763916,
+      "eval_mean_token_accuracy": 0.7912511644738444,
+      "eval_num_tokens": 348955.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 140
+    },
+    {
+      "entropy": 0.7952415809035301,
+      "epoch": 0.38600723763570566,
+      "grad_norm": 0.5387083888053894,
+      "learning_rate": 2.7697012618717034e-05,
+      "loss": 0.740808629989624,
+      "mean_token_accuracy": 0.7941806614398956,
+      "num_tokens": 395798.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38600723763570566,
+      "eval_entropy": 0.7856516191798649,
+      "eval_loss": 0.7324908971786499,
+      "eval_mean_token_accuracy": 0.7980317825011993,
+      "eval_num_tokens": 395798.0,
+      "eval_runtime": 98.6378,
+      "eval_samples_per_second": 14.396,
+      "eval_steps_per_second": 1.805,
+      "step": 160
+    },
+    {
+      "entropy": 0.7537026956677437,
+      "epoch": 0.43425814234016885,
+      "grad_norm": 0.5521425604820251,
+      "learning_rate": 3.118091357704622e-05,
+      "loss": 0.7073616027832031,
+      "mean_token_accuracy": 0.8006261065602303,
+      "num_tokens": 446437.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43425814234016885,
+      "eval_entropy": 0.7470693678668375,
+      "eval_loss": 0.711143434047699,
+      "eval_mean_token_accuracy": 0.801459076029531,
+      "eval_num_tokens": 446437.0,
+      "eval_runtime": 98.6667,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 180
+    },
+    {
+      "entropy": 0.7328549668192863,
+      "epoch": 0.4825090470446321,
+      "grad_norm": 0.4901895821094513,
+      "learning_rate": 3.4664814535375406e-05,
+      "loss": 0.6874266624450683,
+      "mean_token_accuracy": 0.8066758885979652,
+      "num_tokens": 497682.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4825090470446321,
+      "eval_entropy": 0.7374835154983435,
+      "eval_loss": 0.6967977285385132,
+      "eval_mean_token_accuracy": 0.8055616520094068,
+      "eval_num_tokens": 497682.0,
+      "eval_runtime": 98.6234,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 200
+    },
+    {
+      "entropy": 0.7372006759047508,
+      "epoch": 0.5307599517490953,
+      "grad_norm": 0.5516473650932312,
+      "learning_rate": 3.814871549370459e-05,
+      "loss": 0.6906380653381348,
+      "mean_token_accuracy": 0.8057604253292083,
+      "num_tokens": 544670.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5307599517490953,
+      "eval_entropy": 0.710068748573239,
+      "eval_loss": 0.6860950589179993,
+      "eval_mean_token_accuracy": 0.8069355440943429,
+      "eval_num_tokens": 544670.0,
+      "eval_runtime": 98.4618,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 220
+    },
+    {
+      "entropy": 0.7099859565496445,
+      "epoch": 0.5790108564535585,
+      "grad_norm": 0.457378625869751,
+      "learning_rate": 4.1632616452033785e-05,
+      "loss": 0.6674468517303467,
+      "mean_token_accuracy": 0.8087703332304954,
+      "num_tokens": 592926.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5790108564535585,
+      "eval_entropy": 0.717441116156203,
+      "eval_loss": 0.6765118837356567,
+      "eval_mean_token_accuracy": 0.8095786246690857,
+      "eval_num_tokens": 592926.0,
+      "eval_runtime": 98.6828,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 240
+    },
+    {
+      "entropy": 0.7170855522155761,
+      "epoch": 0.6272617611580217,
+      "grad_norm": 0.4604854881763458,
+      "learning_rate": 4.5116517410362974e-05,
+      "loss": 0.6647133827209473,
+      "mean_token_accuracy": 0.8115006521344185,
+      "num_tokens": 639077.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6272617611580217,
+      "eval_entropy": 0.7082702219486237,
+      "eval_loss": 0.6722468733787537,
+      "eval_mean_token_accuracy": 0.8100237367528208,
+      "eval_num_tokens": 639077.0,
+      "eval_runtime": 98.6242,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 260
+    },
+    {
+      "entropy": 0.691286925971508,
+      "epoch": 0.6755126658624849,
+      "grad_norm": 0.4626481831073761,
+      "learning_rate": 4.860041836869215e-05,
+      "loss": 0.6543298721313476,
+      "mean_token_accuracy": 0.8142751976847649,
+      "num_tokens": 691028.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6755126658624849,
+      "eval_entropy": 0.7341568747932992,
+      "eval_loss": 0.658997654914856,
+      "eval_mean_token_accuracy": 0.813439255684949,
+      "eval_num_tokens": 691028.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 280
+    },
+    {
+      "entropy": 0.691513468325138,
+      "epoch": 0.7237635705669482,
+      "grad_norm": 0.5434145331382751,
+      "learning_rate": 5.2084319327021346e-05,
+      "loss": 0.6393396377563476,
+      "mean_token_accuracy": 0.8172704800963402,
+      "num_tokens": 740255.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7237635705669482,
+      "eval_entropy": 0.699160492487168,
+      "eval_loss": 0.6522852182388306,
+      "eval_mean_token_accuracy": 0.8143363323774231,
+      "eval_num_tokens": 740255.0,
+      "eval_runtime": 98.6233,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 300
+    },
+    {
+      "entropy": 0.6918429866433143,
+      "epoch": 0.7720144752714113,
+      "grad_norm": 0.47131890058517456,
+      "learning_rate": 5.556822028535052e-05,
+      "loss": 0.6503581523895263,
+      "mean_token_accuracy": 0.8149650946259499,
+      "num_tokens": 789529.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7720144752714113,
+      "eval_entropy": 0.6949931701247611,
+      "eval_loss": 0.6464414596557617,
+      "eval_mean_token_accuracy": 0.816310279489903,
+      "eval_num_tokens": 789529.0,
+      "eval_runtime": 98.4759,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 320
+    },
+    {
+      "entropy": 0.6926439955830574,
+      "epoch": 0.8202653799758746,
+      "grad_norm": 0.46790942549705505,
+      "learning_rate": 5.905212124367972e-05,
+      "loss": 0.6453651428222656,
+      "mean_token_accuracy": 0.8171535924077034,
+      "num_tokens": 841584.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8202653799758746,
+      "eval_entropy": 0.6810062698433909,
+      "eval_loss": 0.6440736651420593,
+      "eval_mean_token_accuracy": 0.8166696479481258,
+      "eval_num_tokens": 841584.0,
+      "eval_runtime": 98.6479,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 340
+    },
+    {
+      "entropy": 0.7038845136761666,
+      "epoch": 0.8685162846803377,
+      "grad_norm": 0.5038695335388184,
+      "learning_rate": 6.253602220200891e-05,
+      "loss": 0.6558614253997803,
+      "mean_token_accuracy": 0.8127013072371483,
+      "num_tokens": 885769.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8685162846803377,
+      "eval_entropy": 0.6752923710292644,
+      "eval_loss": 0.6410524845123291,
+      "eval_mean_token_accuracy": 0.817534717950928,
+      "eval_num_tokens": 885769.0,
+      "eval_runtime": 98.5986,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 360
+    },
+    {
+      "entropy": 0.6897840186953544,
+      "epoch": 0.916767189384801,
+      "grad_norm": 0.5168097615242004,
+      "learning_rate": 6.601992316033808e-05,
+      "loss": 0.6328047752380371,
+      "mean_token_accuracy": 0.8181755512952804,
+      "num_tokens": 930056.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.916767189384801,
+      "eval_entropy": 0.6816390936964014,
+      "eval_loss": 0.6402509212493896,
+      "eval_mean_token_accuracy": 0.8172605586855599,
+      "eval_num_tokens": 930056.0,
+      "eval_runtime": 98.4787,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 380
+    },
+    {
+      "entropy": 0.6798740655183793,
+      "epoch": 0.9650180940892642,
+      "grad_norm": 0.4609147906303406,
+      "learning_rate": 6.950382411866727e-05,
+      "loss": 0.6358067989349365,
+      "mean_token_accuracy": 0.8193414479494094,
+      "num_tokens": 976892.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9650180940892642,
+      "eval_entropy": 0.6841439577970612,
+      "eval_loss": 0.6335230469703674,
+      "eval_mean_token_accuracy": 0.8190797965848045,
+      "eval_num_tokens": 976892.0,
+      "eval_runtime": 98.5983,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 400
+    },
+    {
+      "entropy": 0.6647876103719076,
+      "epoch": 1.0120627261761157,
+      "grad_norm": 0.46518462896347046,
+      "learning_rate": 7.229074030587624e-05,
+      "loss": 0.6112009525299072,
+      "mean_token_accuracy": 0.8212849283829714,
+      "num_tokens": 1024472.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0120627261761157,
+      "eval_entropy": 0.6612423688508151,
+      "eval_loss": 0.6311497688293457,
+      "eval_mean_token_accuracy": 0.8194578638237514,
+      "eval_num_tokens": 1024472.0,
+      "eval_runtime": 98.4872,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 420
+    },
+    {
+      "entropy": 0.6288895070552826,
+      "epoch": 1.060313630880579,
+      "grad_norm": 0.4342021644115448,
+      "learning_rate": 7.228358026812833e-05,
+      "loss": 0.5853999614715576,
+      "mean_token_accuracy": 0.82901301831007,
+      "num_tokens": 1076964.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060313630880579,
+      "eval_entropy": 0.6567474329069759,
+      "eval_loss": 0.6245927810668945,
+      "eval_mean_token_accuracy": 0.8211770050981072,
+      "eval_num_tokens": 1076964.0,
+      "eval_runtime": 98.6093,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 440
+    },
+    {
+      "entropy": 0.639747454226017,
+      "epoch": 1.1085645355850422,
+      "grad_norm": 0.5338298082351685,
+      "learning_rate": 7.226619357333845e-05,
+      "loss": 0.5847910404205322,
+      "mean_token_accuracy": 0.8279331669211387,
+      "num_tokens": 1126064.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1085645355850422,
+      "eval_entropy": 0.6399250211340658,
+      "eval_loss": 0.6245712637901306,
+      "eval_mean_token_accuracy": 0.8209733875949731,
+      "eval_num_tokens": 1126064.0,
+      "eval_runtime": 98.6581,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 460
+    },
+    {
+      "entropy": 0.6248803421854973,
+      "epoch": 1.1568154402895054,
+      "grad_norm": 0.4402505159378052,
+      "learning_rate": 7.223858514173552e-05,
+      "loss": 0.5847790241241455,
+      "mean_token_accuracy": 0.8273933246731758,
+      "num_tokens": 1176912.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1568154402895054,
+      "eval_entropy": 0.6459465750147787,
+      "eval_loss": 0.615716278553009,
+      "eval_mean_token_accuracy": 0.8228987813665626,
+      "eval_num_tokens": 1176912.0,
+      "eval_runtime": 98.5987,
+      "eval_samples_per_second": 14.402,
+      "eval_steps_per_second": 1.805,
+      "step": 480
+    },
+    {
+      "entropy": 0.6356995061039925,
+      "epoch": 1.2050663449939687,
+      "grad_norm": 0.5318694114685059,
+      "learning_rate": 7.220076278617914e-05,
+      "loss": 0.5902805328369141,
+      "mean_token_accuracy": 0.8279137581586837,
+      "num_tokens": 1223383.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2050663449939687,
+      "eval_entropy": 0.6509443488013879,
+      "eval_loss": 0.6174858808517456,
+      "eval_mean_token_accuracy": 0.8229377031996009,
+      "eval_num_tokens": 1223383.0,
+      "eval_runtime": 98.6156,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 500
+    },
+    {
+      "entropy": 0.6378119193017483,
+      "epoch": 1.2533172496984317,
+      "grad_norm": 0.5385730266571045,
+      "learning_rate": 7.215273720994853e-05,
+      "loss": 0.5850183486938476,
+      "mean_token_accuracy": 0.8282771646976471,
+      "num_tokens": 1273534.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2533172496984317,
+      "eval_entropy": 0.6438969477508845,
+      "eval_loss": 0.6134508848190308,
+      "eval_mean_token_accuracy": 0.823984161186754,
+      "eval_num_tokens": 1273534.0,
+      "eval_runtime": 98.4814,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 520
+    },
+    {
+      "entropy": 0.6330174207687378,
+      "epoch": 1.301568154402895,
+      "grad_norm": 0.4950766861438751,
+      "learning_rate": 7.209452200371367e-05,
+      "loss": 0.5886627674102783,
+      "mean_token_accuracy": 0.828557687997818,
+      "num_tokens": 1323351.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.301568154402895,
+      "eval_entropy": 0.6342737025089478,
+      "eval_loss": 0.611145555973053,
+      "eval_mean_token_accuracy": 0.8250270244780551,
+      "eval_num_tokens": 1323351.0,
+      "eval_runtime": 98.6698,
+      "eval_samples_per_second": 14.391,
+      "eval_steps_per_second": 1.804,
+      "step": 540
+    },
+    {
+      "entropy": 0.6270076125860214,
+      "epoch": 1.3498190591073582,
+      "grad_norm": 0.435604989528656,
+      "learning_rate": 7.20261336416893e-05,
+      "loss": 0.5834732532501221,
+      "mean_token_accuracy": 0.8290600389242172,
+      "num_tokens": 1371508.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3498190591073582,
+      "eval_entropy": 0.6624904408883513,
+      "eval_loss": 0.6063718795776367,
+      "eval_mean_token_accuracy": 0.825260305002834,
+      "eval_num_tokens": 1371508.0,
+      "eval_runtime": 98.6348,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 560
+    },
+    {
+      "entropy": 0.6258987173438072,
+      "epoch": 1.3980699638118215,
+      "grad_norm": 0.45945441722869873,
+      "learning_rate": 7.194759147697293e-05,
+      "loss": 0.5705549716949463,
+      "mean_token_accuracy": 0.8329991281032563,
+      "num_tokens": 1415221.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3980699638118215,
+      "eval_entropy": 0.6500632069753797,
+      "eval_loss": 0.6059838533401489,
+      "eval_mean_token_accuracy": 0.8262855604793249,
+      "eval_num_tokens": 1415221.0,
+      "eval_runtime": 98.6469,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 580
+    },
+    {
+      "entropy": 0.6301489062607288,
+      "epoch": 1.4463208685162847,
+      "grad_norm": 0.5217350721359253,
+      "learning_rate": 7.18589177360681e-05,
+      "loss": 0.5778427124023438,
+      "mean_token_accuracy": 0.8305584967136384,
+      "num_tokens": 1463356.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4463208685162847,
+      "eval_entropy": 0.6478241879618569,
+      "eval_loss": 0.6014655828475952,
+      "eval_mean_token_accuracy": 0.8265940369514937,
+      "eval_num_tokens": 1463356.0,
+      "eval_runtime": 98.6523,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 600
+    },
+    {
+      "entropy": 0.6257719069719314,
+      "epoch": 1.494571773220748,
+      "grad_norm": 0.48310142755508423,
+      "learning_rate": 7.176013751259459e-05,
+      "loss": 0.5713310241699219,
+      "mean_token_accuracy": 0.8305533796548843,
+      "num_tokens": 1512340.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.494571773220748,
+      "eval_entropy": 0.6396401956510008,
+      "eval_loss": 0.5978795289993286,
+      "eval_mean_token_accuracy": 0.8273170630583603,
+      "eval_num_tokens": 1512340.0,
+      "eval_runtime": 98.4958,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 620
+    },
+    {
+      "entropy": 0.6162159442901611,
+      "epoch": 1.5428226779252112,
+      "grad_norm": 0.4500057101249695,
+      "learning_rate": 7.165127876018717e-05,
+      "loss": 0.565293264389038,
+      "mean_token_accuracy": 0.8311659932136536,
+      "num_tokens": 1560911.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5428226779252112,
+      "eval_entropy": 0.6381519051750054,
+      "eval_loss": 0.5963984131813049,
+      "eval_mean_token_accuracy": 0.8268409247478742,
+      "eval_num_tokens": 1560911.0,
+      "eval_runtime": 98.6803,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 640
+    },
+    {
+      "entropy": 0.6088747605681419,
+      "epoch": 1.5910735826296745,
+      "grad_norm": 0.46954667568206787,
+      "learning_rate": 7.153237228458511e-05,
+      "loss": 0.5602522850036621,
+      "mean_token_accuracy": 0.8345131903886795,
+      "num_tokens": 1609409.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5910735826296745,
+      "eval_entropy": 0.6413098875056492,
+      "eval_loss": 0.5938988327980042,
+      "eval_mean_token_accuracy": 0.828771137119679,
+      "eval_num_tokens": 1609409.0,
+      "eval_runtime": 98.4951,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 660
+    },
+    {
+      "entropy": 0.6209868296980858,
+      "epoch": 1.6393244873341375,
+      "grad_norm": 0.3867878317832947,
+      "learning_rate": 7.140345173491446e-05,
+      "loss": 0.5663797378540039,
+      "mean_token_accuracy": 0.8352224260568619,
+      "num_tokens": 1657430.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.6393244873341375,
+      "eval_entropy": 0.6182791748073664,
+      "eval_loss": 0.5919068455696106,
+      "eval_mean_token_accuracy": 0.8286421476455217,
+      "eval_num_tokens": 1657430.0,
+      "eval_runtime": 98.6211,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 680
+    },
+    {
+      "entropy": 0.6191923573613167,
+      "epoch": 1.6875753920386007,
+      "grad_norm": 0.40004226565361023,
+      "learning_rate": 7.12645535941658e-05,
+      "loss": 0.57517991065979,
+      "mean_token_accuracy": 0.8333286643028259,
+      "num_tokens": 1708058.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6875753920386007,
+      "eval_entropy": 0.6244879199212856,
+      "eval_loss": 0.5897895693778992,
+      "eval_mean_token_accuracy": 0.8293621252092083,
+      "eval_num_tokens": 1708058.0,
+      "eval_runtime": 98.4974,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 700
+    },
+    {
+      "entropy": 0.6199282392859459,
+      "epoch": 1.7358262967430638,
+      "grad_norm": 0.36575791239738464,
+      "learning_rate": 7.111571716886997e-05,
+      "loss": 0.5717356681823731,
+      "mean_token_accuracy": 0.8321028783917427,
+      "num_tokens": 1761262.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.7358262967430638,
+      "eval_entropy": 0.6162440127201294,
+      "eval_loss": 0.5864423513412476,
+      "eval_mean_token_accuracy": 0.830874237786518,
+      "eval_num_tokens": 1761262.0,
+      "eval_runtime": 98.6109,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 720
+    },
+    {
+      "entropy": 0.6127455830574036,
+      "epoch": 1.784077201447527,
+      "grad_norm": 0.4778473973274231,
+      "learning_rate": 7.095698457797475e-05,
+      "loss": 0.5664590835571289,
+      "mean_token_accuracy": 0.8324130088090896,
+      "num_tokens": 1812127.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.784077201447527,
+      "eval_entropy": 0.6273092945640006,
+      "eval_loss": 0.5847376585006714,
+      "eval_mean_token_accuracy": 0.831051754817534,
+      "eval_num_tokens": 1812127.0,
+      "eval_runtime": 98.6513,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 740
+    },
+    {
+      "entropy": 0.6141144648194313,
+      "epoch": 1.8323281061519903,
+      "grad_norm": 0.6467089056968689,
+      "learning_rate": 7.078840074092574e-05,
+      "loss": 0.5548617362976074,
+      "mean_token_accuracy": 0.8342376738786698,
+      "num_tokens": 1858403.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.8323281061519903,
+      "eval_entropy": 0.5951976082967908,
+      "eval_loss": 0.5850171446800232,
+      "eval_mean_token_accuracy": 0.8303041870004675,
+      "eval_num_tokens": 1858403.0,
+      "eval_runtime": 98.6601,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 760
+    },
+    {
+      "entropy": 0.5961867064237595,
+      "epoch": 1.8805790108564535,
+      "grad_norm": 0.41280242800712585,
+      "learning_rate": 7.061001336495465e-05,
+      "loss": 0.5511267185211182,
+      "mean_token_accuracy": 0.8392465204000473,
+      "num_tokens": 1907361.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8805790108564535,
+      "eval_entropy": 0.6167017998320333,
+      "eval_loss": 0.5826791524887085,
+      "eval_mean_token_accuracy": 0.8309678050239434,
+      "eval_num_tokens": 1907361.0,
+      "eval_runtime": 98.6564,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 780
+    },
+    {
+      "entropy": 0.6025628983974457,
+      "epoch": 1.9288299155609168,
+      "grad_norm": 0.4078947901725769,
+      "learning_rate": 7.042187293157881e-05,
+      "loss": 0.5552026271820069,
+      "mean_token_accuracy": 0.8359778299927711,
+      "num_tokens": 1956605.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9288299155609168,
+      "eval_entropy": 0.6047008484601974,
+      "eval_loss": 0.5811123847961426,
+      "eval_mean_token_accuracy": 0.8320730804057603,
+      "eval_num_tokens": 1956605.0,
+      "eval_runtime": 98.6033,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 800
+    },
+    {
+      "entropy": 0.6041995719075203,
+      "epoch": 1.97708082026538,
+      "grad_norm": 0.543890118598938,
+      "learning_rate": 7.022403268231535e-05,
+      "loss": 0.554384422302246,
+      "mean_token_accuracy": 0.8380729153752327,
+      "num_tokens": 2003308.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.97708082026538,
+      "eval_entropy": 0.6005127917179901,
+      "eval_loss": 0.5800131559371948,
+      "eval_mean_token_accuracy": 0.8313186547059691,
+      "eval_num_tokens": 2003308.0,
+      "eval_runtime": 98.4119,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 820
+    },
+    {
+      "entropy": 0.567442889396961,
+      "epoch": 2.0241254523522314,
+      "grad_norm": 0.5015465021133423,
+      "learning_rate": 7.001654860361456e-05,
+      "loss": 0.5208251476287842,
+      "mean_token_accuracy": 0.8442784960453327,
+      "num_tokens": 2051331.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0241254523522314,
+      "eval_entropy": 0.5785037809878253,
+      "eval_loss": 0.5865591168403625,
+      "eval_mean_token_accuracy": 0.8307993857378371,
+      "eval_num_tokens": 2051331.0,
+      "eval_runtime": 98.5061,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 840
+    },
+    {
+      "entropy": 0.5552012033760547,
+      "epoch": 2.0723763570566947,
+      "grad_norm": 0.5975943207740784,
+      "learning_rate": 6.979947941101646e-05,
+      "loss": 0.5064489841461182,
+      "mean_token_accuracy": 0.8468396887183189,
+      "num_tokens": 2101528.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0723763570566947,
+      "eval_entropy": 0.5844205239515626,
+      "eval_loss": 0.585284948348999,
+      "eval_mean_token_accuracy": 0.8309797065311604,
+      "eval_num_tokens": 2101528.0,
+      "eval_runtime": 98.4982,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 860
+    },
+    {
+      "entropy": 0.5401638634502888,
+      "epoch": 2.120627261761158,
+      "grad_norm": 0.4194582998752594,
+      "learning_rate": 6.957288653253486e-05,
+      "loss": 0.49037814140319824,
+      "mean_token_accuracy": 0.8528496414422989,
+      "num_tokens": 2149550.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.120627261761158,
+      "eval_entropy": 0.574507936835289,
+      "eval_loss": 0.5836747288703918,
+      "eval_mean_token_accuracy": 0.8325241643391298,
+      "eval_num_tokens": 2149550.0,
+      "eval_runtime": 98.5592,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 880
+    },
+    {
+      "entropy": 0.5606910541653634,
+      "epoch": 2.168878166465621,
+      "grad_norm": 0.5395355820655823,
+      "learning_rate": 6.933683409127401e-05,
+      "loss": 0.5089571475982666,
+      "mean_token_accuracy": 0.8459660410881042,
+      "num_tokens": 2202162.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.168878166465621,
+      "eval_entropy": 0.5779740023479033,
+      "eval_loss": 0.5816043615341187,
+      "eval_mean_token_accuracy": 0.8326318036304431,
+      "eval_num_tokens": 2202162.0,
+      "eval_runtime": 98.4386,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 900
+    },
+    {
+      "entropy": 0.5380903735756875,
+      "epoch": 2.2171290711700844,
+      "grad_norm": 0.46588417887687683,
+      "learning_rate": 6.909138888728257e-05,
+      "loss": 0.4902662754058838,
+      "mean_token_accuracy": 0.8493925213813782,
+      "num_tokens": 2256539.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2171290711700844,
+      "eval_entropy": 0.570154261220707,
+      "eval_loss": 0.5827520489692688,
+      "eval_mean_token_accuracy": 0.8321914167216654,
+      "eval_num_tokens": 2256539.0,
+      "eval_runtime": 98.5674,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 920
+    },
+    {
+      "entropy": 0.558555257320404,
+      "epoch": 2.2653799758745476,
+      "grad_norm": 0.44576048851013184,
+      "learning_rate": 6.883662037864996e-05,
+      "loss": 0.5000699520111084,
+      "mean_token_accuracy": 0.8475695163011551,
+      "num_tokens": 2299698.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2653799758745476,
+      "eval_entropy": 0.5715417749761196,
+      "eval_loss": 0.5831062197685242,
+      "eval_mean_token_accuracy": 0.8327409899636601,
+      "eval_num_tokens": 2299698.0,
+      "eval_runtime": 98.568,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 940
+    },
+    {
+      "entropy": 0.5582522705197335,
+      "epoch": 2.313630880579011,
+      "grad_norm": 0.38966405391693115,
+      "learning_rate": 6.857260066185045e-05,
+      "loss": 0.5132341861724854,
+      "mean_token_accuracy": 0.8466864034533501,
+      "num_tokens": 2347134.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.313630880579011,
+      "eval_entropy": 0.5823918579334624,
+      "eval_loss": 0.5797588229179382,
+      "eval_mean_token_accuracy": 0.8326037566313583,
+      "eval_num_tokens": 2347134.0,
+      "eval_runtime": 98.6222,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 960
+    },
+    {
+      "entropy": 0.5557353265583516,
+      "epoch": 2.361881785283474,
+      "grad_norm": 0.42300811409950256,
+      "learning_rate": 6.829940445134089e-05,
+      "loss": 0.5035375118255615,
+      "mean_token_accuracy": 0.8479988545179367,
+      "num_tokens": 2394568.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.361881785283474,
+      "eval_entropy": 0.5542296482605881,
+      "eval_loss": 0.5800604820251465,
+      "eval_mean_token_accuracy": 0.8336404163515969,
+      "eval_num_tokens": 2394568.0,
+      "eval_runtime": 98.615,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 980
+    },
+    {
+      "entropy": 0.5492161035537719,
+      "epoch": 2.4101326899879374,
+      "grad_norm": 0.47482752799987793,
+      "learning_rate": 6.801710905841724e-05,
+      "loss": 0.5032052040100098,
+      "mean_token_accuracy": 0.8488596931099892,
+      "num_tokens": 2443253.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4101326899879374,
+      "eval_entropy": 0.5755258423558781,
+      "eval_loss": 0.5764774680137634,
+      "eval_mean_token_accuracy": 0.8336571763070781,
+      "eval_num_tokens": 2443253.0,
+      "eval_runtime": 98.4765,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5439704649150372,
+      "epoch": 2.4583835946924006,
+      "grad_norm": 0.3656543493270874,
+      "learning_rate": 6.772579436933644e-05,
+      "loss": 0.49903340339660646,
+      "mean_token_accuracy": 0.8499931409955025,
+      "num_tokens": 2495178.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4583835946924006,
+      "eval_entropy": 0.5712662409530597,
+      "eval_loss": 0.5777108073234558,
+      "eval_mean_token_accuracy": 0.8336941352051296,
+      "eval_num_tokens": 2495178.0,
+      "eval_runtime": 98.678,
+      "eval_samples_per_second": 14.39,
+      "eval_steps_per_second": 1.804,
+      "step": 1020
+    },
+    {
+      "entropy": 0.5448883637785912,
+      "epoch": 2.5066344993968634,
+      "grad_norm": 0.3990268111228943,
+      "learning_rate": 6.742554282270959e-05,
+      "loss": 0.4990490436553955,
+      "mean_token_accuracy": 0.8488436549901962,
+      "num_tokens": 2541733.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.5066344993968634,
+      "eval_entropy": 0.5816385483139017,
+      "eval_loss": 0.5749375820159912,
+      "eval_mean_token_accuracy": 0.833801462074344,
+      "eval_num_tokens": 2541733.0,
+      "eval_runtime": 98.6209,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 1040
+    },
+    {
+      "entropy": 0.5575303137302399,
+      "epoch": 2.554885404101327,
+      "grad_norm": 0.45015832781791687,
+      "learning_rate": 6.711643938617279e-05,
+      "loss": 0.5061063289642334,
+      "mean_token_accuracy": 0.8462719902396202,
+      "num_tokens": 2589479.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.554885404101327,
+      "eval_entropy": 0.5666238901655326,
+      "eval_loss": 0.574820876121521,
+      "eval_mean_token_accuracy": 0.8339746617869045,
+      "eval_num_tokens": 2589479.0,
+      "eval_runtime": 98.5332,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1060
+    },
+    {
+      "entropy": 0.5477629877626896,
+      "epoch": 2.60313630880579,
+      "grad_norm": 0.43583670258522034,
+      "learning_rate": 6.679857153234218e-05,
+      "loss": 0.4998013496398926,
+      "mean_token_accuracy": 0.8465577483177185,
+      "num_tokens": 2639134.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.60313630880579,
+      "eval_entropy": 0.5809191281206152,
+      "eval_loss": 0.5737497210502625,
+      "eval_mean_token_accuracy": 0.8338909467284599,
+      "eval_num_tokens": 2639134.0,
+      "eval_runtime": 98.5093,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5548589795827865,
+      "epoch": 2.651387213510253,
+      "grad_norm": 0.44145169854164124,
+      "learning_rate": 6.647202921406043e-05,
+      "loss": 0.5035099029541016,
+      "mean_token_accuracy": 0.8463503196835518,
+      "num_tokens": 2689324.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651387213510253,
+      "eval_entropy": 0.5626686118292005,
+      "eval_loss": 0.5707846879959106,
+      "eval_mean_token_accuracy": 0.8349729531936432,
+      "eval_num_tokens": 2689324.0,
+      "eval_runtime": 98.473,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5568635657429695,
+      "epoch": 2.6996381182147164,
+      "grad_norm": 0.38482189178466797,
+      "learning_rate": 6.613690483894096e-05,
+      "loss": 0.5090720176696777,
+      "mean_token_accuracy": 0.8481037884950637,
+      "num_tokens": 2736182.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.6996381182147164,
+      "eval_entropy": 0.5752792709998871,
+      "eval_loss": 0.5706034898757935,
+      "eval_mean_token_accuracy": 0.8350465498613507,
+      "eval_num_tokens": 2736182.0,
+      "eval_runtime": 98.6524,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1120
+    },
+    {
+      "entropy": 0.5644363485276699,
+      "epoch": 2.7478890229191797,
+      "grad_norm": 0.4353690445423126,
+      "learning_rate": 6.57932932432178e-05,
+      "loss": 0.5078121185302734,
+      "mean_token_accuracy": 0.846393559873104,
+      "num_tokens": 2781503.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.7478890229191797,
+      "eval_entropy": 0.5670730096570561,
+      "eval_loss": 0.5699080228805542,
+      "eval_mean_token_accuracy": 0.8354134288396728,
+      "eval_num_tokens": 2781503.0,
+      "eval_runtime": 98.5697,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1140
+    },
+    {
+      "entropy": 0.547992604970932,
+      "epoch": 2.796139927623643,
+      "grad_norm": 0.4126238524913788,
+      "learning_rate": 6.544129166490802e-05,
+      "loss": 0.4943515777587891,
+      "mean_token_accuracy": 0.8496095523238182,
+      "num_tokens": 2831126.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796139927623643,
+      "eval_entropy": 0.5552005295673114,
+      "eval_loss": 0.5671029090881348,
+      "eval_mean_token_accuracy": 0.8364275894807965,
+      "eval_num_tokens": 2831126.0,
+      "eval_runtime": 98.5412,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1160
+    },
+    {
+      "entropy": 0.561221782118082,
+      "epoch": 2.844390832328106,
+      "grad_norm": 0.47753170132637024,
+      "learning_rate": 6.508099971629446e-05,
+      "loss": 0.5094280242919922,
+      "mean_token_accuracy": 0.8462468951940536,
+      "num_tokens": 2878560.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.844390832328106,
+      "eval_entropy": 0.5532578974962234,
+      "eval_loss": 0.5714129209518433,
+      "eval_mean_token_accuracy": 0.8357450737712089,
+      "eval_num_tokens": 2878560.0,
+      "eval_runtime": 98.5358,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5400963060557842,
+      "epoch": 2.8926417370325694,
+      "grad_norm": 0.38486021757125854,
+      "learning_rate": 6.471251935573679e-05,
+      "loss": 0.49443836212158204,
+      "mean_token_accuracy": 0.8484652638435364,
+      "num_tokens": 2927569.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8926417370325694,
+      "eval_entropy": 0.5697383115465722,
+      "eval_loss": 0.5681865215301514,
+      "eval_mean_token_accuracy": 0.8350923453154189,
+      "eval_num_tokens": 2927569.0,
+      "eval_runtime": 98.4136,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5381059750914574,
+      "epoch": 2.9408926417370327,
+      "grad_norm": 0.49116969108581543,
+      "learning_rate": 6.43359548588183e-05,
+      "loss": 0.4878206729888916,
+      "mean_token_accuracy": 0.8522022038698196,
+      "num_tokens": 2978577.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9408926417370327,
+      "eval_entropy": 0.5614734472853414,
+      "eval_loss": 0.5665943622589111,
+      "eval_mean_token_accuracy": 0.8360684296388304,
+      "eval_num_tokens": 2978577.0,
+      "eval_runtime": 98.5682,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1220
+    },
+    {
+      "entropy": 0.5569418989121914,
+      "epoch": 2.989143546441496,
+      "grad_norm": 0.39628803730010986,
+      "learning_rate": 6.39514127888373e-05,
+      "loss": 0.5087827682495117,
+      "mean_token_accuracy": 0.8476181954145432,
+      "num_tokens": 3025565.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.989143546441496,
+      "eval_entropy": 0.5740904657358534,
+      "eval_loss": 0.5637108683586121,
+      "eval_mean_token_accuracy": 0.8366153638684348,
+      "eval_num_tokens": 3025565.0,
+      "eval_runtime": 98.4904,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5067426371268737,
+      "epoch": 3.0361881785283473,
+      "grad_norm": 0.4593591094017029,
+      "learning_rate": 6.355900196665094e-05,
+      "loss": 0.4515994071960449,
+      "mean_token_accuracy": 0.8621168640943674,
+      "num_tokens": 3073705.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.0361881785283473,
+      "eval_entropy": 0.5176989454566763,
+      "eval_loss": 0.5825926661491394,
+      "eval_mean_token_accuracy": 0.8353930775369152,
+      "eval_num_tokens": 3073705.0,
+      "eval_runtime": 98.5654,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 1260
+    },
+    {
+      "entropy": 0.4712781891226768,
+      "epoch": 3.0844390832328106,
+      "grad_norm": 0.46717822551727295,
+      "learning_rate": 6.315883343988008e-05,
+      "loss": 0.4228676795959473,
+      "mean_token_accuracy": 0.8672123551368713,
+      "num_tokens": 3126555.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.0844390832328106,
+      "eval_entropy": 0.5229383730486538,
+      "eval_loss": 0.5875843167304993,
+      "eval_mean_token_accuracy": 0.8347370524084969,
+      "eval_num_tokens": 3126555.0,
+      "eval_runtime": 98.5459,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1280
+    },
+    {
+      "entropy": 0.48235836550593375,
+      "epoch": 3.132689987937274,
+      "grad_norm": 0.5073364973068237,
+      "learning_rate": 6.275102045148425e-05,
+      "loss": 0.4365375518798828,
+      "mean_token_accuracy": 0.8644895792007447,
+      "num_tokens": 3179060.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.132689987937274,
+      "eval_entropy": 0.5151654947339819,
+      "eval_loss": 0.5826129913330078,
+      "eval_mean_token_accuracy": 0.8359152366606037,
+      "eval_num_tokens": 3179060.0,
+      "eval_runtime": 98.3713,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 1300
+    },
+    {
+      "entropy": 0.4756111137568951,
+      "epoch": 3.180940892641737,
+      "grad_norm": 0.4896017909049988,
+      "learning_rate": 6.233567840771526e-05,
+      "loss": 0.43177642822265627,
+      "mean_token_accuracy": 0.8645799145102501,
+      "num_tokens": 3234212.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.180940892641737,
+      "eval_entropy": 0.5227621813838401,
+      "eval_loss": 0.5796895027160645,
+      "eval_mean_token_accuracy": 0.8355028448479899,
+      "eval_num_tokens": 3234212.0,
+      "eval_runtime": 98.5538,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1320
+    },
+    {
+      "entropy": 0.47541345208883284,
+      "epoch": 3.2291917973462003,
+      "grad_norm": 0.5064273476600647,
+      "learning_rate": 6.191292484545842e-05,
+      "loss": 0.4289750099182129,
+      "mean_token_accuracy": 0.8659383848309516,
+      "num_tokens": 3282544.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2291917973462003,
+      "eval_entropy": 0.5286844237132019,
+      "eval_loss": 0.579947829246521,
+      "eval_mean_token_accuracy": 0.8353635648663125,
+      "eval_num_tokens": 3282544.0,
+      "eval_runtime": 98.589,
+      "eval_samples_per_second": 14.403,
+      "eval_steps_per_second": 1.805,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4916382558643818,
+      "epoch": 3.2774427020506636,
+      "grad_norm": 0.5225116014480591,
+      "learning_rate": 6.148287939897118e-05,
+      "loss": 0.44352192878723146,
+      "mean_token_accuracy": 0.8630174115300179,
+      "num_tokens": 3332532.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2774427020506636,
+      "eval_entropy": 0.5264859509267165,
+      "eval_loss": 0.5774893164634705,
+      "eval_mean_token_accuracy": 0.8363003827882617,
+      "eval_num_tokens": 3332532.0,
+      "eval_runtime": 98.5858,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1360
+    },
+    {
+      "entropy": 0.4990046709775925,
+      "epoch": 3.325693606755127,
+      "grad_norm": 0.548435628414154,
+      "learning_rate": 6.104566376602795e-05,
+      "loss": 0.4463675498962402,
+      "mean_token_accuracy": 0.8630342438817025,
+      "num_tokens": 3378178.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.325693606755127,
+      "eval_entropy": 0.522659856616781,
+      "eval_loss": 0.5790286660194397,
+      "eval_mean_token_accuracy": 0.8355394277010071,
+      "eval_num_tokens": 3378178.0,
+      "eval_runtime": 98.4707,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1380
+    },
+    {
+      "entropy": 0.48943497613072395,
+      "epoch": 3.37394451145959,
+      "grad_norm": 0.6222821474075317,
+      "learning_rate": 6.060140167348109e-05,
+      "loss": 0.4356812000274658,
+      "mean_token_accuracy": 0.8644352272152901,
+      "num_tokens": 3425739.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.37394451145959,
+      "eval_entropy": 0.5065617703654793,
+      "eval_loss": 0.5842992067337036,
+      "eval_mean_token_accuracy": 0.8358401662847968,
+      "eval_num_tokens": 3425739.0,
+      "eval_runtime": 98.4775,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 1400
+    },
+    {
+      "entropy": 0.47648413479328156,
+      "epoch": 3.422195416164053,
+      "grad_norm": 0.5207330584526062,
+      "learning_rate": 6.015021884224766e-05,
+      "loss": 0.42724952697753904,
+      "mean_token_accuracy": 0.8663465172052384,
+      "num_tokens": 3472737.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.422195416164053,
+      "eval_entropy": 0.5232780023237292,
+      "eval_loss": 0.5769163966178894,
+      "eval_mean_token_accuracy": 0.8356800042511372,
+      "eval_num_tokens": 3472737.0,
+      "eval_runtime": 98.5005,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 1420
+    },
+    {
+      "entropy": 0.4767202906310558,
+      "epoch": 3.470446320868516,
+      "grad_norm": 0.4695364534854889,
+      "learning_rate": 5.969224295173193e-05,
+      "loss": 0.42672252655029297,
+      "mean_token_accuracy": 0.8657647699117661,
+      "num_tokens": 3519802.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.470446320868516,
+      "eval_entropy": 0.5146279760291067,
+      "eval_loss": 0.577904224395752,
+      "eval_mean_token_accuracy": 0.8365512393163831,
+      "eval_num_tokens": 3519802.0,
+      "eval_runtime": 98.5825,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4855267658829689,
+      "epoch": 3.5186972255729794,
+      "grad_norm": 0.6333373785018921,
+      "learning_rate": 5.92276036036936e-05,
+      "loss": 0.4310962677001953,
+      "mean_token_accuracy": 0.864738130569458,
+      "num_tokens": 3568814.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.5186972255729794,
+      "eval_entropy": 0.5260473481389913,
+      "eval_loss": 0.5770025253295898,
+      "eval_mean_token_accuracy": 0.8357346339172191,
+      "eval_num_tokens": 3568814.0,
+      "eval_runtime": 98.462,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 1460
+    },
+    {
+      "entropy": 0.4917050123214722,
+      "epoch": 3.5669481302774426,
+      "grad_norm": 0.5833342671394348,
+      "learning_rate": 5.875643228557199e-05,
+      "loss": 0.44533047676086424,
+      "mean_token_accuracy": 0.8642907112836837,
+      "num_tokens": 3616502.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.5669481302774426,
+      "eval_entropy": 0.5278292167722509,
+      "eval_loss": 0.5773910880088806,
+      "eval_mean_token_accuracy": 0.83580608548743,
+      "eval_num_tokens": 3616502.0,
+      "eval_runtime": 98.4817,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5025837384164333,
+      "epoch": 3.615199034981906,
+      "grad_norm": 0.5484253168106079,
+      "learning_rate": 5.827886233327671e-05,
+      "loss": 0.44471564292907717,
+      "mean_token_accuracy": 0.8624475196003913,
+      "num_tokens": 3663750.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.615199034981906,
+      "eval_entropy": 0.5118448720889145,
+      "eval_loss": 0.5786464810371399,
+      "eval_mean_token_accuracy": 0.8370849305994055,
+      "eval_num_tokens": 3663750.0,
+      "eval_runtime": 98.5208,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 1500
+    },
+    {
+      "entropy": 0.494297581911087,
+      "epoch": 3.663449939686369,
+      "grad_norm": 0.5475512146949768,
+      "learning_rate": 5.7795028893455165e-05,
+      "loss": 0.44118366241455076,
+      "mean_token_accuracy": 0.8630154177546501,
+      "num_tokens": 3709337.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663449939686369,
+      "eval_entropy": 0.5250470112548785,
+      "eval_loss": 0.5746602416038513,
+      "eval_mean_token_accuracy": 0.8370710711800651,
+      "eval_num_tokens": 3709337.0,
+      "eval_runtime": 98.54,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4840923048555851,
+      "epoch": 3.7117008443908324,
+      "grad_norm": 0.39450952410697937,
+      "learning_rate": 5.730506888524761e-05,
+      "loss": 0.443526554107666,
+      "mean_token_accuracy": 0.8627525731921196,
+      "num_tokens": 3759111.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.7117008443908324,
+      "eval_entropy": 0.5305671740448876,
+      "eval_loss": 0.5697008371353149,
+      "eval_mean_token_accuracy": 0.837059904015466,
+      "eval_num_tokens": 3759111.0,
+      "eval_runtime": 98.6288,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 1540
+    },
+    {
+      "entropy": 0.49613324701786043,
+      "epoch": 3.7599517490952956,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 5.680912096154071e-05,
+      "loss": 0.44313778877258303,
+      "mean_token_accuracy": 0.8614323407411575,
+      "num_tokens": 3806869.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7599517490952956,
+      "eval_entropy": 0.522762700580479,
+      "eval_loss": 0.571703314781189,
+      "eval_mean_token_accuracy": 0.8381530970669864,
+      "eval_num_tokens": 3806869.0,
+      "eval_runtime": 98.6493,
+      "eval_samples_per_second": 14.394,
+      "eval_steps_per_second": 1.804,
+      "step": 1560
+    },
+    {
+      "entropy": 0.4979200787842274,
+      "epoch": 3.808202653799759,
+      "grad_norm": 0.5275628566741943,
+      "learning_rate": 5.630732546973038e-05,
+      "loss": 0.4462272167205811,
+      "mean_token_accuracy": 0.8615201056003571,
+      "num_tokens": 3854699.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808202653799759,
+      "eval_entropy": 0.5232290047942922,
+      "eval_loss": 0.568954586982727,
+      "eval_mean_token_accuracy": 0.8380421128835571,
+      "eval_num_tokens": 3854699.0,
+      "eval_runtime": 98.484,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 1580
+    },
+    {
+      "entropy": 0.4957688055932522,
+      "epoch": 3.856453558504222,
+      "grad_norm": 0.5806206464767456,
+      "learning_rate": 5.5799824412005126e-05,
+      "loss": 0.44518418312072755,
+      "mean_token_accuracy": 0.860895662009716,
+      "num_tokens": 3903750.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856453558504222,
+      "eval_entropy": 0.5238539929805177,
+      "eval_loss": 0.573081910610199,
+      "eval_mean_token_accuracy": 0.8369396288743179,
+      "eval_num_tokens": 3903750.0,
+      "eval_runtime": 98.5731,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5005358196794987,
+      "epoch": 3.9047044632086854,
+      "grad_norm": 0.5114495754241943,
+      "learning_rate": 5.528676140516111e-05,
+      "loss": 0.45078649520874026,
+      "mean_token_accuracy": 0.860628567636013,
+      "num_tokens": 3952913.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9047044632086854,
+      "eval_entropy": 0.5245393164707034,
+      "eval_loss": 0.5683825612068176,
+      "eval_mean_token_accuracy": 0.8380641599049728,
+      "eval_num_tokens": 3952913.0,
+      "eval_runtime": 98.5275,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1620
+    },
+    {
+      "entropy": 0.4839292488992214,
+      "epoch": 3.952955367913148,
+      "grad_norm": 0.6169933080673218,
+      "learning_rate": 5.4768281639960344e-05,
+      "loss": 0.43407382965087893,
+      "mean_token_accuracy": 0.867341585457325,
+      "num_tokens": 4002571.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.952955367913148,
+      "eval_entropy": 0.5277225765284528,
+      "eval_loss": 0.567344069480896,
+      "eval_mean_token_accuracy": 0.8386145431673928,
+      "eval_num_tokens": 4002571.0,
+      "eval_runtime": 98.5702,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 1640
+    },
+    {
+      "entropy": 0.4952052327302786,
+      "epoch": 4.0,
+      "grad_norm": 2.080777168273926,
+      "learning_rate": 5.424453184004336e-05,
+      "loss": 0.4525449752807617,
+      "mean_token_accuracy": 0.8622837005517422,
+      "num_tokens": 4049116.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5197472232446242,
+      "eval_loss": 0.57077956199646,
+      "eval_mean_token_accuracy": 0.8381094560864266,
+      "eval_num_tokens": 4049116.0,
+      "eval_runtime": 98.5498,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 1660
+    },
+    {
+      "entropy": 0.42416453957557676,
+      "epoch": 4.048250904704463,
+      "grad_norm": 0.41175922751426697,
+      "learning_rate": 5.3715660220408146e-05,
+      "loss": 0.3610882043838501,
+      "mean_token_accuracy": 0.8832842499017716,
+      "num_tokens": 4097986.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048250904704463,
+      "eval_entropy": 0.4814340365066957,
+      "eval_loss": 0.5963674783706665,
+      "eval_mean_token_accuracy": 0.8364584405770462,
+      "eval_num_tokens": 4097986.0,
+      "eval_runtime": 98.4464,
+      "eval_samples_per_second": 14.424,
+      "eval_steps_per_second": 1.808,
+      "step": 1680
+    },
+    {
+      "entropy": 0.4193890765309334,
+      "epoch": 4.0965018094089265,
+      "grad_norm": 0.4773254990577698,
+      "learning_rate": 5.3181816445467126e-05,
+      "loss": 0.36023797988891604,
+      "mean_token_accuracy": 0.884614959359169,
+      "num_tokens": 4147967.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0965018094089265,
+      "eval_entropy": 0.47795870448096417,
+      "eval_loss": 0.6029345393180847,
+      "eval_mean_token_accuracy": 0.8354013039824668,
+      "eval_num_tokens": 4147967.0,
+      "eval_runtime": 98.5784,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 1700
+    },
+    {
+      "entropy": 0.41848619729280473,
+      "epoch": 4.144752714113389,
+      "grad_norm": 0.7293462157249451,
+      "learning_rate": 5.264315158669384e-05,
+      "loss": 0.36621780395507814,
+      "mean_token_accuracy": 0.882377204298973,
+      "num_tokens": 4198623.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.144752714113389,
+      "eval_entropy": 0.4683974619661824,
+      "eval_loss": 0.6010014414787292,
+      "eval_mean_token_accuracy": 0.8358984790491254,
+      "eval_num_tokens": 4198623.0,
+      "eval_runtime": 98.5406,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1720
+    },
+    {
+      "entropy": 0.41377718672156333,
+      "epoch": 4.193003618817853,
+      "grad_norm": 0.5061255693435669,
+      "learning_rate": 5.209981807987152e-05,
+      "loss": 0.36024787425994875,
+      "mean_token_accuracy": 0.8844522640109063,
+      "num_tokens": 4248406.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193003618817853,
+      "eval_entropy": 0.4666026450609893,
+      "eval_loss": 0.6050319075584412,
+      "eval_mean_token_accuracy": 0.8357557922266843,
+      "eval_num_tokens": 4248406.0,
+      "eval_runtime": 98.6056,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4185848042368889,
+      "epoch": 4.241254523522316,
+      "grad_norm": 0.49589282274246216,
+      "learning_rate": 5.155196968195561e-05,
+      "loss": 0.36044313907623293,
+      "mean_token_accuracy": 0.8863146558403969,
+      "num_tokens": 4296608.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.241254523522316,
+      "eval_entropy": 0.4660757115382827,
+      "eval_loss": 0.6089843511581421,
+      "eval_mean_token_accuracy": 0.8352006492989786,
+      "eval_num_tokens": 4296608.0,
+      "eval_runtime": 98.5425,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4249664135277271,
+      "epoch": 4.2895054282267795,
+      "grad_norm": 0.6652432680130005,
+      "learning_rate": 5.099976142756227e-05,
+      "loss": 0.37266385555267334,
+      "mean_token_accuracy": 0.880594964325428,
+      "num_tokens": 4345389.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.2895054282267795,
+      "eval_entropy": 0.48157174483443915,
+      "eval_loss": 0.5987246036529541,
+      "eval_mean_token_accuracy": 0.8361060595914219,
+      "eval_num_tokens": 4345389.0,
+      "eval_runtime": 98.4652,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4195803053677082,
+      "epoch": 4.337756332931242,
+      "grad_norm": 0.6217629313468933,
+      "learning_rate": 5.044334958509551e-05,
+      "loss": 0.3646178483963013,
+      "mean_token_accuracy": 0.8808894723653793,
+      "num_tokens": 4395878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337756332931242,
+      "eval_entropy": 0.47762840182593697,
+      "eval_loss": 0.5994009375572205,
+      "eval_mean_token_accuracy": 0.8366789422678144,
+      "eval_num_tokens": 4395878.0,
+      "eval_runtime": 98.6677,
+      "eval_samples_per_second": 14.392,
+      "eval_steps_per_second": 1.804,
+      "step": 1800
+    },
+    {
+      "entropy": 0.42473565861582757,
+      "epoch": 4.386007237635706,
+      "grad_norm": 0.6135475635528564,
+      "learning_rate": 4.9882891612525023e-05,
+      "loss": 0.37249753475189207,
+      "mean_token_accuracy": 0.8800962314009666,
+      "num_tokens": 4440908.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386007237635706,
+      "eval_entropy": 0.47359335924802204,
+      "eval_loss": 0.6024664640426636,
+      "eval_mean_token_accuracy": 0.8355274846714534,
+      "eval_num_tokens": 4440908.0,
+      "eval_runtime": 98.5103,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 1820
+    },
+    {
+      "entropy": 0.4212189674377441,
+      "epoch": 4.434258142340169,
+      "grad_norm": 0.5874977707862854,
+      "learning_rate": 4.931854611282748e-05,
+      "loss": 0.36766781806945803,
+      "mean_token_accuracy": 0.8837772741913795,
+      "num_tokens": 4489671.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434258142340169,
+      "eval_entropy": 0.46670635798004234,
+      "eval_loss": 0.6085236668586731,
+      "eval_mean_token_accuracy": 0.8355892652206207,
+      "eval_num_tokens": 4489671.0,
+      "eval_runtime": 98.5871,
+      "eval_samples_per_second": 14.404,
+      "eval_steps_per_second": 1.806,
+      "step": 1840
+    },
+    {
+      "entropy": 0.41604730263352396,
+      "epoch": 4.4825090470446325,
+      "grad_norm": 0.5763360261917114,
+      "learning_rate": 4.875047278910371e-05,
+      "loss": 0.36643099784851074,
+      "mean_token_accuracy": 0.8825299561023712,
+      "num_tokens": 4541286.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4825090470446325,
+      "eval_entropy": 0.47700903121005284,
+      "eval_loss": 0.606238842010498,
+      "eval_mean_token_accuracy": 0.8348092550642034,
+      "eval_num_tokens": 4541286.0,
+      "eval_runtime": 98.5305,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1860
+    },
+    {
+      "entropy": 0.4189125746488571,
+      "epoch": 4.530759951749095,
+      "grad_norm": 0.5916087627410889,
+      "learning_rate": 4.817883239938471e-05,
+      "loss": 0.3703711748123169,
+      "mean_token_accuracy": 0.8825044095516205,
+      "num_tokens": 4589986.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.530759951749095,
+      "eval_entropy": 0.47982015703501324,
+      "eval_loss": 0.6003754138946533,
+      "eval_mean_token_accuracy": 0.8351969742373134,
+      "eval_num_tokens": 4589986.0,
+      "eval_runtime": 98.4111,
+      "eval_samples_per_second": 14.429,
+      "eval_steps_per_second": 1.809,
+      "step": 1880
+    },
+    {
+      "entropy": 0.4429341182112694,
+      "epoch": 4.579010856453559,
+      "grad_norm": 0.6441562175750732,
+      "learning_rate": 4.7603786711138866e-05,
+      "loss": 0.3824815511703491,
+      "mean_token_accuracy": 0.8784194126725197,
+      "num_tokens": 4636548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579010856453559,
+      "eval_entropy": 0.4740966230296017,
+      "eval_loss": 0.5988267660140991,
+      "eval_mean_token_accuracy": 0.8369577316085944,
+      "eval_num_tokens": 4636548.0,
+      "eval_runtime": 98.5291,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41137003526091576,
+      "epoch": 4.627261761158022,
+      "grad_norm": 0.6663628816604614,
+      "learning_rate": 4.702549845549385e-05,
+      "loss": 0.36178064346313477,
+      "mean_token_accuracy": 0.8830229461193084,
+      "num_tokens": 4686918.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627261761158022,
+      "eval_entropy": 0.4730259802904022,
+      "eval_loss": 0.6001402139663696,
+      "eval_mean_token_accuracy": 0.8366120139534554,
+      "eval_num_tokens": 4686918.0,
+      "eval_runtime": 98.5556,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 1920
+    },
+    {
+      "entropy": 0.4264552935957909,
+      "epoch": 4.675512665862485,
+      "grad_norm": 0.6708512902259827,
+      "learning_rate": 4.64441312811855e-05,
+      "loss": 0.376303505897522,
+      "mean_token_accuracy": 0.8826978862285614,
+      "num_tokens": 4734637.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675512665862485,
+      "eval_entropy": 0.4687648815720269,
+      "eval_loss": 0.5982367992401123,
+      "eval_mean_token_accuracy": 0.8368918728292658,
+      "eval_num_tokens": 4734637.0,
+      "eval_runtime": 98.4928,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4378764644265175,
+      "epoch": 4.723763570566948,
+      "grad_norm": 0.7098956108093262,
+      "learning_rate": 4.585984970824725e-05,
+      "loss": 0.38264145851135256,
+      "mean_token_accuracy": 0.8768767058849335,
+      "num_tokens": 4783109.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.723763570566948,
+      "eval_entropy": 0.46603381248672354,
+      "eval_loss": 0.6010599732398987,
+      "eval_mean_token_accuracy": 0.8361037319965576,
+      "eval_num_tokens": 4783109.0,
+      "eval_runtime": 98.5157,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 1960
+    },
+    {
+      "entropy": 0.44009693413972856,
+      "epoch": 4.772014475271411,
+      "grad_norm": 0.7529758214950562,
+      "learning_rate": 4.527281908145284e-05,
+      "loss": 0.379723048210144,
+      "mean_token_accuracy": 0.8799900904297828,
+      "num_tokens": 4828531.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.772014475271411,
+      "eval_entropy": 0.47042709873633437,
+      "eval_loss": 0.5984888672828674,
+      "eval_mean_token_accuracy": 0.8375813492228475,
+      "eval_num_tokens": 4828531.0,
+      "eval_runtime": 98.3352,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 1980
+    },
+    {
+      "entropy": 0.43963938802480695,
+      "epoch": 4.820265379975875,
+      "grad_norm": 0.530246376991272,
+      "learning_rate": 4.468320552352577e-05,
+      "loss": 0.3855365037918091,
+      "mean_token_accuracy": 0.8766040325164794,
+      "num_tokens": 4877038.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.820265379975875,
+      "eval_entropy": 0.4773505919435051,
+      "eval_loss": 0.5956445932388306,
+      "eval_mean_token_accuracy": 0.8365653522228926,
+      "eval_num_tokens": 4877038.0,
+      "eval_runtime": 98.6015,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 2000
+    },
+    {
+      "entropy": 0.4241503834724426,
+      "epoch": 4.868516284680338,
+      "grad_norm": 0.5264645218849182,
+      "learning_rate": 4.409117588812845e-05,
+      "loss": 0.3725905418395996,
+      "mean_token_accuracy": 0.8816388875246048,
+      "num_tokens": 4928583.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.868516284680338,
+      "eval_entropy": 0.4738255136468437,
+      "eval_loss": 0.5987666845321655,
+      "eval_mean_token_accuracy": 0.8362959995028678,
+      "eval_num_tokens": 4928583.0,
+      "eval_runtime": 98.6141,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 1.805,
+      "step": 2020
+    },
+    {
+      "entropy": 0.4451489016413689,
+      "epoch": 4.916767189384801,
+      "grad_norm": 0.5322316884994507,
+      "learning_rate": 4.3496897712644684e-05,
+      "loss": 0.3903519630432129,
+      "mean_token_accuracy": 0.8752380400896073,
+      "num_tokens": 4975223.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.916767189384801,
+      "eval_entropy": 0.4662560015247109,
+      "eval_loss": 0.5963740944862366,
+      "eval_mean_token_accuracy": 0.8378168083308788,
+      "eval_num_tokens": 4975223.0,
+      "eval_runtime": 98.6613,
+      "eval_samples_per_second": 14.393,
+      "eval_steps_per_second": 1.804,
+      "step": 2040
+    },
+    {
+      "entropy": 0.43658976554870604,
+      "epoch": 4.965018094089264,
+      "grad_norm": 0.6989460587501526,
+      "learning_rate": 4.29005391707685e-05,
+      "loss": 0.3851702451705933,
+      "mean_token_accuracy": 0.8790383815765381,
+      "num_tokens": 5021142.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.965018094089264,
+      "eval_entropy": 0.4791082282749455,
+      "eval_loss": 0.5912667512893677,
+      "eval_mean_token_accuracy": 0.83739292286755,
+      "eval_num_tokens": 5021142.0,
+      "eval_runtime": 98.5606,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 2060
+    },
+    {
+      "entropy": 0.40598615316244274,
+      "epoch": 5.012062726176116,
+      "grad_norm": 0.5303810238838196,
+      "learning_rate": 4.2302269024913e-05,
+      "loss": 0.3450628280639648,
+      "mean_token_accuracy": 0.8893755567379487,
+      "num_tokens": 5073226.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.012062726176116,
+      "eval_entropy": 0.42836171187711564,
+      "eval_loss": 0.6335396766662598,
+      "eval_mean_token_accuracy": 0.8349329397249757,
+      "eval_num_tokens": 5073226.0,
+      "eval_runtime": 98.3503,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2080
+    },
+    {
+      "entropy": 0.3517624609172344,
+      "epoch": 5.060313630880579,
+      "grad_norm": 0.6402326226234436,
+      "learning_rate": 4.170225657845261e-05,
+      "loss": 0.2898552894592285,
+      "mean_token_accuracy": 0.904600378870964,
+      "num_tokens": 5120052.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.060313630880579,
+      "eval_entropy": 0.4094204810563098,
+      "eval_loss": 0.652080237865448,
+      "eval_mean_token_accuracy": 0.8331165705504042,
+      "eval_num_tokens": 5120052.0,
+      "eval_runtime": 98.3573,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2100
+    },
+    {
+      "entropy": 0.3560003057122231,
+      "epoch": 5.108564535585042,
+      "grad_norm": 0.9274800419807434,
+      "learning_rate": 4.110067162781221e-05,
+      "loss": 0.2926265954971313,
+      "mean_token_accuracy": 0.9040415585041046,
+      "num_tokens": 5165984.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.108564535585042,
+      "eval_entropy": 0.41318756994906436,
+      "eval_loss": 0.6567395925521851,
+      "eval_mean_token_accuracy": 0.83362964197491,
+      "eval_num_tokens": 5165984.0,
+      "eval_runtime": 98.5042,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2120
+    },
+    {
+      "entropy": 0.35080533251166346,
+      "epoch": 5.156815440289505,
+      "grad_norm": 0.6713427901268005,
+      "learning_rate": 4.049768441441677e-05,
+      "loss": 0.29822506904602053,
+      "mean_token_accuracy": 0.9027505293488503,
+      "num_tokens": 5214442.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.156815440289505,
+      "eval_entropy": 0.416848015416874,
+      "eval_loss": 0.6563553214073181,
+      "eval_mean_token_accuracy": 0.8333374677079447,
+      "eval_num_tokens": 5214442.0,
+      "eval_runtime": 98.4261,
+      "eval_samples_per_second": 14.427,
+      "eval_steps_per_second": 1.808,
+      "step": 2140
+    },
+    {
+      "entropy": 0.3736354000866413,
+      "epoch": 5.205066344993969,
+      "grad_norm": 0.5667551755905151,
+      "learning_rate": 3.989346557651503e-05,
+      "loss": 0.3088909387588501,
+      "mean_token_accuracy": 0.9012129113078118,
+      "num_tokens": 5260853.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.205066344993969,
+      "eval_entropy": 0.41974830309326727,
+      "eval_loss": 0.6504881381988525,
+      "eval_mean_token_accuracy": 0.8349264759695931,
+      "eval_num_tokens": 5260853.0,
+      "eval_runtime": 98.5007,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2160
+    },
+    {
+      "entropy": 0.35063633099198344,
+      "epoch": 5.253317249698432,
+      "grad_norm": 0.7737690210342407,
+      "learning_rate": 3.9288186100890903e-05,
+      "loss": 0.2888066053390503,
+      "mean_token_accuracy": 0.9040960729122162,
+      "num_tokens": 5312338.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.253317249698432,
+      "eval_entropy": 0.4058936465992017,
+      "eval_loss": 0.6578952670097351,
+      "eval_mean_token_accuracy": 0.8352396407823884,
+      "eval_num_tokens": 5312338.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2180
+    },
+    {
+      "entropy": 0.3545838728547096,
+      "epoch": 5.301568154402895,
+      "grad_norm": 0.8293929100036621,
+      "learning_rate": 3.8682017274476215e-05,
+      "loss": 0.29347708225250246,
+      "mean_token_accuracy": 0.902638104557991,
+      "num_tokens": 5359039.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.301568154402895,
+      "eval_entropy": 0.41738414345832353,
+      "eval_loss": 0.6543260812759399,
+      "eval_mean_token_accuracy": 0.8336448260907376,
+      "eval_num_tokens": 5359039.0,
+      "eval_runtime": 98.499,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2200
+    },
+    {
+      "entropy": 0.3416903533041477,
+      "epoch": 5.349819059107358,
+      "grad_norm": 0.7580088973045349,
+      "learning_rate": 3.8075130635878556e-05,
+      "loss": 0.28760058879852296,
+      "mean_token_accuracy": 0.906610831618309,
+      "num_tokens": 5409807.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.349819059107358,
+      "eval_entropy": 0.4121073748958245,
+      "eval_loss": 0.6563934087753296,
+      "eval_mean_token_accuracy": 0.833708616120092,
+      "eval_num_tokens": 5409807.0,
+      "eval_runtime": 98.3808,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2220
+    },
+    {
+      "entropy": 0.352034630253911,
+      "epoch": 5.398069963811821,
+      "grad_norm": 0.7154259085655212,
+      "learning_rate": 3.746769792683787e-05,
+      "loss": 0.2967940092086792,
+      "mean_token_accuracy": 0.9023159235715866,
+      "num_tokens": 5458308.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.398069963811821,
+      "eval_entropy": 0.4149914874454563,
+      "eval_loss": 0.6444063186645508,
+      "eval_mean_token_accuracy": 0.8360218017288809,
+      "eval_num_tokens": 5458308.0,
+      "eval_runtime": 98.4832,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2240
+    },
+    {
+      "entropy": 0.3539439059793949,
+      "epoch": 5.446320868516285,
+      "grad_norm": 0.8078773617744446,
+      "learning_rate": 3.685989104362558e-05,
+      "loss": 0.2928986310958862,
+      "mean_token_accuracy": 0.905618867278099,
+      "num_tokens": 5505380.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.446320868516285,
+      "eval_entropy": 0.4099769044793054,
+      "eval_loss": 0.6588519215583801,
+      "eval_mean_token_accuracy": 0.8353862765799748,
+      "eval_num_tokens": 5505380.0,
+      "eval_runtime": 98.3781,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2260
+    },
+    {
+      "entropy": 0.35821759328246117,
+      "epoch": 5.4945717732207475,
+      "grad_norm": 0.8760809302330017,
+      "learning_rate": 3.6251881988399994e-05,
+      "loss": 0.29706525802612305,
+      "mean_token_accuracy": 0.9018172711133957,
+      "num_tokens": 5553385.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.4945717732207475,
+      "eval_entropy": 0.40791591667057425,
+      "eval_loss": 0.6583756804466248,
+      "eval_mean_token_accuracy": 0.8349186125096311,
+      "eval_num_tokens": 5553385.0,
+      "eval_runtime": 98.4808,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2280
+    },
+    {
+      "entropy": 0.34672123566269875,
+      "epoch": 5.542822677925211,
+      "grad_norm": 0.8768810033798218,
+      "learning_rate": 3.564384282053165e-05,
+      "loss": 0.2885677576065063,
+      "mean_token_accuracy": 0.9063566774129868,
+      "num_tokens": 5604413.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.542822677925211,
+      "eval_entropy": 0.412923522377282,
+      "eval_loss": 0.6535082459449768,
+      "eval_mean_token_accuracy": 0.8350685978873392,
+      "eval_num_tokens": 5604413.0,
+      "eval_runtime": 98.3842,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2300
+    },
+    {
+      "entropy": 0.35606732666492463,
+      "epoch": 5.591073582629674,
+      "grad_norm": 0.7237412929534912,
+      "learning_rate": 3.503594560791267e-05,
+      "loss": 0.30418403148651124,
+      "mean_token_accuracy": 0.9004160806536674,
+      "num_tokens": 5657218.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.591073582629674,
+      "eval_entropy": 0.42160589155856143,
+      "eval_loss": 0.6412338018417358,
+      "eval_mean_token_accuracy": 0.834980196162556,
+      "eval_num_tokens": 5657218.0,
+      "eval_runtime": 98.3674,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2320
+    },
+    {
+      "entropy": 0.35330103561282156,
+      "epoch": 5.639324487334138,
+      "grad_norm": 0.5864736437797546,
+      "learning_rate": 3.442836237826349e-05,
+      "loss": 0.29608047008514404,
+      "mean_token_accuracy": 0.9029312193393707,
+      "num_tokens": 5709140.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.639324487334138,
+      "eval_entropy": 0.4273778943533308,
+      "eval_loss": 0.6407549977302551,
+      "eval_mean_token_accuracy": 0.8351332397943132,
+      "eval_num_tokens": 5709140.0,
+      "eval_runtime": 98.5533,
+      "eval_samples_per_second": 14.408,
+      "eval_steps_per_second": 1.806,
+      "step": 2340
+    },
+    {
+      "entropy": 0.353700140863657,
+      "epoch": 5.6875753920386005,
+      "grad_norm": 0.744305431842804,
+      "learning_rate": 3.382126507045109e-05,
+      "loss": 0.3004793643951416,
+      "mean_token_accuracy": 0.9015571728348732,
+      "num_tokens": 5760973.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.6875753920386005,
+      "eval_entropy": 0.4262272163053577,
+      "eval_loss": 0.6443405151367188,
+      "eval_mean_token_accuracy": 0.8354629797212193,
+      "eval_num_tokens": 5760973.0,
+      "eval_runtime": 98.3719,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2360
+    },
+    {
+      "entropy": 0.36056690439581873,
+      "epoch": 5.735826296743064,
+      "grad_norm": 0.6552110910415649,
+      "learning_rate": 3.321482548583228e-05,
+      "loss": 0.2950709104537964,
+      "mean_token_accuracy": 0.9033273816108703,
+      "num_tokens": 5808058.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.735826296743064,
+      "eval_entropy": 0.41902332275771026,
+      "eval_loss": 0.6472867727279663,
+      "eval_mean_token_accuracy": 0.8354150207524889,
+      "eval_num_tokens": 5808058.0,
+      "eval_runtime": 98.3956,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2380
+    },
+    {
+      "entropy": 0.35097662582993505,
+      "epoch": 5.784077201447527,
+      "grad_norm": 0.670831024646759,
+      "learning_rate": 3.260921523963606e-05,
+      "loss": 0.29224560260772703,
+      "mean_token_accuracy": 0.9042141750454903,
+      "num_tokens": 5857700.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.784077201447527,
+      "eval_entropy": 0.4284254325909561,
+      "eval_loss": 0.6407246589660645,
+      "eval_mean_token_accuracy": 0.8345498831084605,
+      "eval_num_tokens": 5857700.0,
+      "eval_runtime": 98.3651,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2400
+    },
+    {
+      "entropy": 0.35357567593455314,
+      "epoch": 5.832328106151991,
+      "grad_norm": 0.8287181258201599,
+      "learning_rate": 3.200460571239839e-05,
+      "loss": 0.2948127746582031,
+      "mean_token_accuracy": 0.9042218685150146,
+      "num_tokens": 5907064.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.832328106151991,
+      "eval_entropy": 0.4094803942723221,
+      "eval_loss": 0.6497060060501099,
+      "eval_mean_token_accuracy": 0.8358801147910986,
+      "eval_num_tokens": 5907064.0,
+      "eval_runtime": 98.3681,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 2420
+    },
+    {
+      "entropy": 0.36021770909428596,
+      "epoch": 5.8805790108564535,
+      "grad_norm": 0.6800532937049866,
+      "learning_rate": 3.140116800146367e-05,
+      "loss": 0.30136709213256835,
+      "mean_token_accuracy": 0.9020729199051857,
+      "num_tokens": 5954440.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.8805790108564535,
+      "eval_entropy": 0.4202987858083811,
+      "eval_loss": 0.6428810954093933,
+      "eval_mean_token_accuracy": 0.8355459637856215,
+      "eval_num_tokens": 5954440.0,
+      "eval_runtime": 98.5147,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2440
+    },
+    {
+      "entropy": 0.34706664383411406,
+      "epoch": 5.928829915560916,
+      "grad_norm": 0.6911236643791199,
+      "learning_rate": 3.07990728725661e-05,
+      "loss": 0.28953113555908205,
+      "mean_token_accuracy": 0.9055185422301293,
+      "num_tokens": 6002089.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.928829915560916,
+      "eval_entropy": 0.406153663825453,
+      "eval_loss": 0.6541930437088013,
+      "eval_mean_token_accuracy": 0.834371502814668,
+      "eval_num_tokens": 6002089.0,
+      "eval_runtime": 98.3494,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 2460
+    },
+    {
+      "entropy": 0.34852043315768244,
+      "epoch": 5.97708082026538,
+      "grad_norm": 0.7544664144515991,
+      "learning_rate": 3.0198490711505038e-05,
+      "loss": 0.28950369358062744,
+      "mean_token_accuracy": 0.9057677388191223,
+      "num_tokens": 6051074.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.97708082026538,
+      "eval_entropy": 0.4174931427735961,
+      "eval_loss": 0.6448875665664673,
+      "eval_mean_token_accuracy": 0.8358313615402478,
+      "eval_num_tokens": 6051074.0,
+      "eval_runtime": 98.4786,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2480
+    },
+    {
+      "entropy": 0.31319292042499935,
+      "epoch": 6.024125452352232,
+      "grad_norm": 0.7616425156593323,
+      "learning_rate": 2.9599591475927893e-05,
+      "loss": 0.24788825511932372,
+      "mean_token_accuracy": 0.9198053983541635,
+      "num_tokens": 6101526.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.024125452352232,
+      "eval_entropy": 0.3601247634780541,
+      "eval_loss": 0.7335909008979797,
+      "eval_mean_token_accuracy": 0.8309332149082356,
+      "eval_num_tokens": 6101526.0,
+      "eval_runtime": 98.3569,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 2500
+    },
+    {
+      "entropy": 0.27905523739755156,
+      "epoch": 6.072376357056695,
+      "grad_norm": 0.7366685271263123,
+      "learning_rate": 2.9002544647234015e-05,
+      "loss": 0.21279211044311525,
+      "mean_token_accuracy": 0.929847776889801,
+      "num_tokens": 6147939.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.072376357056695,
+      "eval_entropy": 0.36974305969275784,
+      "eval_loss": 0.7080109715461731,
+      "eval_mean_token_accuracy": 0.8330447111236915,
+      "eval_num_tokens": 6147939.0,
+      "eval_runtime": 98.4908,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2520
+    },
+    {
+      "entropy": 0.2782863024622202,
+      "epoch": 6.120627261761158,
+      "grad_norm": 0.9167613387107849,
+      "learning_rate": 2.840751918261356e-05,
+      "loss": 0.20897085666656495,
+      "mean_token_accuracy": 0.9294204145669938,
+      "num_tokens": 6193631.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.120627261761158,
+      "eval_entropy": 0.366466105319141,
+      "eval_loss": 0.7170661091804504,
+      "eval_mean_token_accuracy": 0.8328661680891273,
+      "eval_num_tokens": 6193631.0,
+      "eval_runtime": 98.4385,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2540
+    },
+    {
+      "entropy": 0.28355731777846815,
+      "epoch": 6.168878166465621,
+      "grad_norm": 0.8010057806968689,
+      "learning_rate": 2.7814683467234538e-05,
+      "loss": 0.211204195022583,
+      "mean_token_accuracy": 0.9297966167330742,
+      "num_tokens": 6240258.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.168878166465621,
+      "eval_entropy": 0.36755596988656547,
+      "eval_loss": 0.7274795174598694,
+      "eval_mean_token_accuracy": 0.8303602230682802,
+      "eval_num_tokens": 6240258.0,
+      "eval_runtime": 98.487,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2560
+    },
+    {
+      "entropy": 0.27822012901306153,
+      "epoch": 6.217129071170085,
+      "grad_norm": 0.7516391277313232,
+      "learning_rate": 2.7224205266591812e-05,
+      "loss": 0.2117089033126831,
+      "mean_token_accuracy": 0.9294167309999466,
+      "num_tokens": 6289137.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.217129071170085,
+      "eval_entropy": 0.37035224410925016,
+      "eval_loss": 0.7123831510543823,
+      "eval_mean_token_accuracy": 0.8318134563022785,
+      "eval_num_tokens": 6289137.0,
+      "eval_runtime": 98.5029,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2580
+    },
+    {
+      "entropy": 0.2784307342022657,
+      "epoch": 6.265379975874548,
+      "grad_norm": 0.8481637239456177,
+      "learning_rate": 2.6636251679031508e-05,
+      "loss": 0.21476459503173828,
+      "mean_token_accuracy": 0.9293668136000633,
+      "num_tokens": 6339215.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.265379975874548,
+      "eval_entropy": 0.3675622491354353,
+      "eval_loss": 0.7221675515174866,
+      "eval_mean_token_accuracy": 0.8314910557832611,
+      "eval_num_tokens": 6339215.0,
+      "eval_runtime": 98.3892,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 2600
+    },
+    {
+      "entropy": 0.27380565330386164,
+      "epoch": 6.3136308805790105,
+      "grad_norm": 0.7563941478729248,
+      "learning_rate": 2.605098908846406e-05,
+      "loss": 0.2117525815963745,
+      "mean_token_accuracy": 0.9299218639731407,
+      "num_tokens": 6390057.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.3136308805790105,
+      "eval_entropy": 0.3676499978879864,
+      "eval_loss": 0.7249981164932251,
+      "eval_mean_token_accuracy": 0.8315788115678209,
+      "eval_num_tokens": 6390057.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 2620
+    },
+    {
+      "entropy": 0.26819119676947595,
+      "epoch": 6.361881785283474,
+      "grad_norm": 0.9161278605461121,
+      "learning_rate": 2.546858311727963e-05,
+      "loss": 0.2020425319671631,
+      "mean_token_accuracy": 0.9322459831833839,
+      "num_tokens": 6443300.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.361881785283474,
+      "eval_entropy": 0.3589264298422953,
+      "eval_loss": 0.737293004989624,
+      "eval_mean_token_accuracy": 0.8307386482699534,
+      "eval_num_tokens": 6443300.0,
+      "eval_runtime": 98.4767,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2640
+    },
+    {
+      "entropy": 0.275980044901371,
+      "epoch": 6.410132689987937,
+      "grad_norm": 0.7255368828773499,
+      "learning_rate": 2.488919857947885e-05,
+      "loss": 0.21207733154296876,
+      "mean_token_accuracy": 0.9280627354979515,
+      "num_tokens": 6491445.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.410132689987937,
+      "eval_entropy": 0.3729423410102223,
+      "eval_loss": 0.7137707471847534,
+      "eval_mean_token_accuracy": 0.8315478007445175,
+      "eval_num_tokens": 6491445.0,
+      "eval_runtime": 98.384,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2660
+    },
+    {
+      "entropy": 0.2794003043323755,
+      "epoch": 6.458383594692401,
+      "grad_norm": 0.817340612411499,
+      "learning_rate": 2.4312999434032465e-05,
+      "loss": 0.2155231714248657,
+      "mean_token_accuracy": 0.930096672475338,
+      "num_tokens": 6541352.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.458383594692401,
+      "eval_entropy": 0.36857131709543506,
+      "eval_loss": 0.7273393869400024,
+      "eval_mean_token_accuracy": 0.8302122903673836,
+      "eval_num_tokens": 6541352.0,
+      "eval_runtime": 98.5119,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 2680
+    },
+    {
+      "entropy": 0.28550401143729687,
+      "epoch": 6.5066344993968634,
+      "grad_norm": 0.7945013642311096,
+      "learning_rate": 2.374014873848286e-05,
+      "loss": 0.21428213119506836,
+      "mean_token_accuracy": 0.9295192703604698,
+      "num_tokens": 6589480.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.5066344993968634,
+      "eval_entropy": 0.3720237272843886,
+      "eval_loss": 0.7190349698066711,
+      "eval_mean_token_accuracy": 0.8307539365264807,
+      "eval_num_tokens": 6589480.0,
+      "eval_runtime": 98.3972,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2700
+    },
+    {
+      "entropy": 0.2789166793227196,
+      "epoch": 6.554885404101327,
+      "grad_norm": 0.7820216417312622,
+      "learning_rate": 2.3170808602800694e-05,
+      "loss": 0.21065421104431153,
+      "mean_token_accuracy": 0.9293019056320191,
+      "num_tokens": 6635214.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.554885404101327,
+      "eval_entropy": 0.36198576715555086,
+      "eval_loss": 0.7250529527664185,
+      "eval_mean_token_accuracy": 0.8310985079642093,
+      "eval_num_tokens": 6635214.0,
+      "eval_runtime": 98.516,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 2720
+    },
+    {
+      "entropy": 0.2805891681462526,
+      "epoch": 6.60313630880579,
+      "grad_norm": 0.8589067459106445,
+      "learning_rate": 2.2605140143509734e-05,
+      "loss": 0.21532771587371827,
+      "mean_token_accuracy": 0.9301707550883294,
+      "num_tokens": 6682916.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.60313630880579,
+      "eval_entropy": 0.36839111149311066,
+      "eval_loss": 0.720038890838623,
+      "eval_mean_token_accuracy": 0.8314180384191234,
+      "eval_num_tokens": 6682916.0,
+      "eval_runtime": 98.3864,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2740
+    },
+    {
+      "entropy": 0.2923452850431204,
+      "epoch": 6.651387213510254,
+      "grad_norm": 1.1464570760726929,
+      "learning_rate": 2.2043303438092746e-05,
+      "loss": 0.22156639099121095,
+      "mean_token_accuracy": 0.9259255439043045,
+      "num_tokens": 6728644.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.651387213510254,
+      "eval_entropy": 0.36308020152402726,
+      "eval_loss": 0.7283808588981628,
+      "eval_mean_token_accuracy": 0.8307237183110098,
+      "eval_num_tokens": 6728644.0,
+      "eval_runtime": 98.3733,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 2760
+    },
+    {
+      "entropy": 0.2749974336475134,
+      "epoch": 6.699638118214716,
+      "grad_norm": 1.0338026285171509,
+      "learning_rate": 2.1485457479691453e-05,
+      "loss": 0.2111821413040161,
+      "mean_token_accuracy": 0.9308380514383316,
+      "num_tokens": 6776170.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.699638118214716,
+      "eval_entropy": 0.35679193979568696,
+      "eval_loss": 0.7316031455993652,
+      "eval_mean_token_accuracy": 0.8316076066386834,
+      "eval_num_tokens": 6776170.0,
+      "eval_runtime": 98.4012,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 2780
+    },
+    {
+      "entropy": 0.2827217370271683,
+      "epoch": 6.74788902291918,
+      "grad_norm": 0.6860962510108948,
+      "learning_rate": 2.0931760132113435e-05,
+      "loss": 0.219130802154541,
+      "mean_token_accuracy": 0.9272046208381652,
+      "num_tokens": 6823320.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.74788902291918,
+      "eval_entropy": 0.367448085479522,
+      "eval_loss": 0.7210394144058228,
+      "eval_mean_token_accuracy": 0.8324049028118005,
+      "eval_num_tokens": 6823320.0,
+      "eval_runtime": 98.3867,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 2800
+    },
+    {
+      "entropy": 0.28139964006841184,
+      "epoch": 6.796139927623643,
+      "grad_norm": 0.7198498249053955,
+      "learning_rate": 2.038236808515841e-05,
+      "loss": 0.2179172992706299,
+      "mean_token_accuracy": 0.9282332047820091,
+      "num_tokens": 6874282.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.796139927623643,
+      "eval_entropy": 0.3658287344353922,
+      "eval_loss": 0.7204482555389404,
+      "eval_mean_token_accuracy": 0.8324238242727987,
+      "eval_num_tokens": 6874282.0,
+      "eval_runtime": 98.3796,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2820
+    },
+    {
+      "entropy": 0.28035391345620153,
+      "epoch": 6.844390832328106,
+      "grad_norm": 0.8470256924629211,
+      "learning_rate": 1.9837436810277026e-05,
+      "loss": 0.21534478664398193,
+      "mean_token_accuracy": 0.9284071952104569,
+      "num_tokens": 6923379.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.844390832328106,
+      "eval_entropy": 0.3615414013018769,
+      "eval_loss": 0.7223248481750488,
+      "eval_mean_token_accuracy": 0.8328172581249409,
+      "eval_num_tokens": 6923379.0,
+      "eval_runtime": 98.4426,
+      "eval_samples_per_second": 14.425,
+      "eval_steps_per_second": 1.808,
+      "step": 2840
+    },
+    {
+      "entropy": 0.27447608187794686,
+      "epoch": 6.892641737032569,
+      "grad_norm": 1.1274205446243286,
+      "learning_rate": 1.9297120516574132e-05,
+      "loss": 0.2151768684387207,
+      "mean_token_accuracy": 0.9284873619675637,
+      "num_tokens": 6971734.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.892641737032569,
+      "eval_entropy": 0.36038200892089456,
+      "eval_loss": 0.7199941873550415,
+      "eval_mean_token_accuracy": 0.8334562922461649,
+      "eval_num_tokens": 6971734.0,
+      "eval_runtime": 98.4807,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 2860
+    },
+    {
+      "entropy": 0.27447244003415106,
+      "epoch": 6.940892641737032,
+      "grad_norm": 0.857891857624054,
+      "learning_rate": 1.876157210716952e-05,
+      "loss": 0.21279585361480713,
+      "mean_token_accuracy": 0.9296008720993996,
+      "num_tokens": 7024380.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.940892641737032,
+      "eval_entropy": 0.3578391455197602,
+      "eval_loss": 0.7355349063873291,
+      "eval_mean_token_accuracy": 0.8319483108065101,
+      "eval_num_tokens": 7024380.0,
+      "eval_runtime": 98.3412,
+      "eval_samples_per_second": 14.44,
+      "eval_steps_per_second": 1.81,
+      "step": 2880
+    },
+    {
+      "entropy": 0.27788355611264703,
+      "epoch": 6.989143546441496,
+      "grad_norm": 0.8024702668190002,
+      "learning_rate": 1.8230943135928133e-05,
+      "loss": 0.21420254707336425,
+      "mean_token_accuracy": 0.9291928946971894,
+      "num_tokens": 7074034.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.989143546441496,
+      "eval_entropy": 0.35886382102296593,
+      "eval_loss": 0.7313408851623535,
+      "eval_mean_token_accuracy": 0.8322128275137269,
+      "eval_num_tokens": 7074034.0,
+      "eval_runtime": 98.4564,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 2900
+    },
+    {
+      "entropy": 0.24298458527296019,
+      "epoch": 7.036188178528348,
+      "grad_norm": 0.8567366600036621,
+      "learning_rate": 1.7705383764572093e-05,
+      "loss": 0.15583932399749756,
+      "mean_token_accuracy": 0.9481525069628006,
+      "num_tokens": 7122753.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.036188178528348,
+      "eval_entropy": 0.3236005037520709,
+      "eval_loss": 0.8240236043930054,
+      "eval_mean_token_accuracy": 0.8270553162928378,
+      "eval_num_tokens": 7122753.0,
+      "eval_runtime": 98.3819,
+      "eval_samples_per_second": 14.434,
+      "eval_steps_per_second": 1.809,
+      "step": 2920
+    },
+    {
+      "entropy": 0.2195398263633251,
+      "epoch": 7.084439083232811,
+      "grad_norm": 0.8163788914680481,
+      "learning_rate": 1.7185042720186728e-05,
+      "loss": 0.14468494653701783,
+      "mean_token_accuracy": 0.9525706857442856,
+      "num_tokens": 7171505.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.084439083232811,
+      "eval_entropy": 0.32686926424503326,
+      "eval_loss": 0.8029564023017883,
+      "eval_mean_token_accuracy": 0.8280681498934713,
+      "eval_num_tokens": 7171505.0,
+      "eval_runtime": 98.7239,
+      "eval_samples_per_second": 14.384,
+      "eval_steps_per_second": 1.803,
+      "step": 2940
+    },
+    {
+      "entropy": 0.21137732304632664,
+      "epoch": 7.132689987937274,
+      "grad_norm": 0.7098442316055298,
+      "learning_rate": 1.6670067253132565e-05,
+      "loss": 0.1382979393005371,
+      "mean_token_accuracy": 0.9557823866605759,
+      "num_tokens": 7219913.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.132689987937274,
+      "eval_entropy": 0.32376345288887454,
+      "eval_loss": 0.8175393342971802,
+      "eval_mean_token_accuracy": 0.8281539598207796,
+      "eval_num_tokens": 7219913.0,
+      "eval_runtime": 98.4728,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 2960
+    },
+    {
+      "entropy": 0.21736963428556919,
+      "epoch": 7.180940892641737,
+      "grad_norm": 0.8196840286254883,
+      "learning_rate": 1.616060309537518e-05,
+      "loss": 0.14342793226242065,
+      "mean_token_accuracy": 0.9531146377325058,
+      "num_tokens": 7265991.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.180940892641737,
+      "eval_entropy": 0.3240856607810835,
+      "eval_loss": 0.816273033618927,
+      "eval_mean_token_accuracy": 0.8283853041991759,
+      "eval_num_tokens": 7265991.0,
+      "eval_runtime": 98.4989,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 2980
+    },
+    {
+      "entropy": 0.2128997717052698,
+      "epoch": 7.2291917973462,
+      "grad_norm": 0.7581701278686523,
+      "learning_rate": 1.5656794419244872e-05,
+      "loss": 0.14148796796798707,
+      "mean_token_accuracy": 0.9551757454872132,
+      "num_tokens": 7317184.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.2291917973462,
+      "eval_entropy": 0.3229605887043342,
+      "eval_loss": 0.8140764832496643,
+      "eval_mean_token_accuracy": 0.8288295952121864,
+      "eval_num_tokens": 7317184.0,
+      "eval_runtime": 98.4981,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3000
+    },
+    {
+      "entropy": 0.214377661049366,
+      "epoch": 7.277442702050664,
+      "grad_norm": 0.6986305117607117,
+      "learning_rate": 1.5158783796637473e-05,
+      "loss": 0.14369987249374389,
+      "mean_token_accuracy": 0.9528444185853004,
+      "num_tokens": 7365364.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.277442702050664,
+      "eval_entropy": 0.32001526101251665,
+      "eval_loss": 0.8190910816192627,
+      "eval_mean_token_accuracy": 0.829054100459881,
+      "eval_num_tokens": 7365364.0,
+      "eval_runtime": 98.5783,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3020
+    },
+    {
+      "entropy": 0.20811213366687298,
+      "epoch": 7.325693606755126,
+      "grad_norm": 0.9613916277885437,
+      "learning_rate": 1.4666712158668313e-05,
+      "loss": 0.14105329513549805,
+      "mean_token_accuracy": 0.9532626882195473,
+      "num_tokens": 7415795.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.325693606755126,
+      "eval_entropy": 0.31941057892327895,
+      "eval_loss": 0.8219555616378784,
+      "eval_mean_token_accuracy": 0.8282468245270547,
+      "eval_num_tokens": 7415795.0,
+      "eval_runtime": 98.5199,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3040
+    },
+    {
+      "entropy": 0.2161025382578373,
+      "epoch": 7.37394451145959,
+      "grad_norm": 0.802244246006012,
+      "learning_rate": 1.418071875579028e-05,
+      "loss": 0.14670108556747435,
+      "mean_token_accuracy": 0.9525101512670517,
+      "num_tokens": 7464083.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.37394451145959,
+      "eval_entropy": 0.3249504163526417,
+      "eval_loss": 0.8168519139289856,
+      "eval_mean_token_accuracy": 0.8281265336476015,
+      "eval_num_tokens": 7464083.0,
+      "eval_runtime": 98.4837,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 3060
+    },
+    {
+      "entropy": 0.21994274333119393,
+      "epoch": 7.422195416164053,
+      "grad_norm": 0.847614586353302,
+      "learning_rate": 1.370094111838762e-05,
+      "loss": 0.14627058506011964,
+      "mean_token_accuracy": 0.9523571625351905,
+      "num_tokens": 7513568.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.422195416164053,
+      "eval_entropy": 0.3224444846423824,
+      "eval_loss": 0.8206745982170105,
+      "eval_mean_token_accuracy": 0.8285097687431936,
+      "eval_num_tokens": 7513568.0,
+      "eval_runtime": 98.5699,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3080
+    },
+    {
+      "entropy": 0.21474989503622055,
+      "epoch": 7.470446320868517,
+      "grad_norm": 0.8615112900733948,
+      "learning_rate": 1.3227515017856523e-05,
+      "loss": 0.14166855812072754,
+      "mean_token_accuracy": 0.9545643255114555,
+      "num_tokens": 7560503.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.470446320868517,
+      "eval_entropy": 0.3131480763634939,
+      "eval_loss": 0.8318397402763367,
+      "eval_mean_token_accuracy": 0.8293128348468395,
+      "eval_num_tokens": 7560503.0,
+      "eval_runtime": 98.5748,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3100
+    },
+    {
+      "entropy": 0.20732707269489764,
+      "epoch": 7.518697225572979,
+      "grad_norm": 0.8595603108406067,
+      "learning_rate": 1.2760574428183327e-05,
+      "loss": 0.1390355110168457,
+      "mean_token_accuracy": 0.9545612305402755,
+      "num_tokens": 7610472.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.518697225572979,
+      "eval_entropy": 0.32303216425555475,
+      "eval_loss": 0.8206163048744202,
+      "eval_mean_token_accuracy": 0.8276272057817223,
+      "eval_num_tokens": 7610472.0,
+      "eval_runtime": 98.6296,
+      "eval_samples_per_second": 14.397,
+      "eval_steps_per_second": 1.805,
+      "step": 3120
+    },
+    {
+      "entropy": 0.20272110737860202,
+      "epoch": 7.566948130277443,
+      "grad_norm": 0.9171860814094543,
+      "learning_rate": 1.2300251488031559e-05,
+      "loss": 0.13609397411346436,
+      "mean_token_accuracy": 0.9557161509990693,
+      "num_tokens": 7662595.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.566948130277443,
+      "eval_entropy": 0.31589331360680334,
+      "eval_loss": 0.8248094916343689,
+      "eval_mean_token_accuracy": 0.8292221159077762,
+      "eval_num_tokens": 7662595.0,
+      "eval_runtime": 98.4574,
+      "eval_samples_per_second": 14.422,
+      "eval_steps_per_second": 1.808,
+      "step": 3140
+    },
+    {
+      "entropy": 0.21678446158766745,
+      "epoch": 7.615199034981906,
+      "grad_norm": 0.870278000831604,
+      "learning_rate": 1.1846676463348161e-05,
+      "loss": 0.14677901268005372,
+      "mean_token_accuracy": 0.9534812569618225,
+      "num_tokens": 7711331.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.615199034981906,
+      "eval_entropy": 0.3183298571391052,
+      "eval_loss": 0.8235951662063599,
+      "eval_mean_token_accuracy": 0.8291523583819357,
+      "eval_num_tokens": 7711331.0,
+      "eval_runtime": 98.5743,
+      "eval_samples_per_second": 14.405,
+      "eval_steps_per_second": 1.806,
+      "step": 3160
+    },
+    {
+      "entropy": 0.2208289708942175,
+      "epoch": 7.66344993968637,
+      "grad_norm": 0.8425227999687195,
+      "learning_rate": 1.1399977710499864e-05,
+      "loss": 0.146539568901062,
+      "mean_token_accuracy": 0.9524248003959656,
+      "num_tokens": 7758759.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.66344993968637,
+      "eval_entropy": 0.3229933256346188,
+      "eval_loss": 0.8207331299781799,
+      "eval_mean_token_accuracy": 0.8284426217668512,
+      "eval_num_tokens": 7758759.0,
+      "eval_runtime": 98.6951,
+      "eval_samples_per_second": 14.388,
+      "eval_steps_per_second": 1.804,
+      "step": 3180
+    },
+    {
+      "entropy": 0.2222116433084011,
+      "epoch": 7.711700844390832,
+      "grad_norm": 0.8922525644302368,
+      "learning_rate": 1.0960281639949764e-05,
+      "loss": 0.14516948461532592,
+      "mean_token_accuracy": 0.952321757376194,
+      "num_tokens": 7805791.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.711700844390832,
+      "eval_entropy": 0.32119952912410993,
+      "eval_loss": 0.8213328123092651,
+      "eval_mean_token_accuracy": 0.8285770188556628,
+      "eval_num_tokens": 7805791.0,
+      "eval_runtime": 98.6437,
+      "eval_samples_per_second": 14.395,
+      "eval_steps_per_second": 1.804,
+      "step": 3200
+    },
+    {
+      "entropy": 0.21560499407351016,
+      "epoch": 7.759951749095295,
+      "grad_norm": 0.9485028386116028,
+      "learning_rate": 1.0527712680484752e-05,
+      "loss": 0.1465426802635193,
+      "mean_token_accuracy": 0.9517758965492249,
+      "num_tokens": 7854924.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.759951749095295,
+      "eval_entropy": 0.3180267366298129,
+      "eval_loss": 0.8266745805740356,
+      "eval_mean_token_accuracy": 0.8283248446630628,
+      "eval_num_tokens": 7854924.0,
+      "eval_runtime": 98.5477,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3220
+    },
+    {
+      "entropy": 0.20934636779129506,
+      "epoch": 7.808202653799759,
+      "grad_norm": 0.7967023253440857,
+      "learning_rate": 1.0102393244003528e-05,
+      "loss": 0.14084668159484864,
+      "mean_token_accuracy": 0.9531109407544136,
+      "num_tokens": 7905721.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.808202653799759,
+      "eval_entropy": 0.31779277040047593,
+      "eval_loss": 0.8262729048728943,
+      "eval_mean_token_accuracy": 0.8288095492995187,
+      "eval_num_tokens": 7905721.0,
+      "eval_runtime": 98.6036,
+      "eval_samples_per_second": 14.401,
+      "eval_steps_per_second": 1.805,
+      "step": 3240
+    },
+    {
+      "entropy": 0.215614727884531,
+      "epoch": 7.856453558504222,
+      "grad_norm": 0.7962915301322937,
+      "learning_rate": 9.684443690875602e-06,
+      "loss": 0.14136850833892822,
+      "mean_token_accuracy": 0.9531293407082557,
+      "num_tokens": 7955010.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.856453558504222,
+      "eval_entropy": 0.31656171481930806,
+      "eval_loss": 0.8303171992301941,
+      "eval_mean_token_accuracy": 0.8285069612974532,
+      "eval_num_tokens": 7955010.0,
+      "eval_runtime": 98.5105,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3260
+    },
+    {
+      "entropy": 0.21908088140189647,
+      "epoch": 7.904704463208685,
+      "grad_norm": 0.9550889134407043,
+      "learning_rate": 9.273982295880618e-06,
+      "loss": 0.14732669591903685,
+      "mean_token_accuracy": 0.9522618219256401,
+      "num_tokens": 8001797.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.904704463208685,
+      "eval_entropy": 0.3193875088283185,
+      "eval_loss": 0.8240101933479309,
+      "eval_mean_token_accuracy": 0.8285468552219734,
+      "eval_num_tokens": 8001797.0,
+      "eval_runtime": 98.4506,
+      "eval_samples_per_second": 14.423,
+      "eval_steps_per_second": 1.808,
+      "step": 3280
+    },
+    {
+      "entropy": 0.21472855806350707,
+      "epoch": 7.952955367913148,
+      "grad_norm": 0.8032136559486389,
+      "learning_rate": 8.871125214738089e-06,
+      "loss": 0.1441211223602295,
+      "mean_token_accuracy": 0.9532426640391349,
+      "num_tokens": 8050462.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.952955367913148,
+      "eval_entropy": 0.3145652976766061,
+      "eval_loss": 0.8368120789527893,
+      "eval_mean_token_accuracy": 0.8285725685987579,
+      "eval_num_tokens": 8050462.0,
+      "eval_runtime": 98.5667,
+      "eval_samples_per_second": 14.406,
+      "eval_steps_per_second": 1.806,
+      "step": 3300
+    },
+    {
+      "entropy": 0.20777193781657097,
+      "epoch": 8.0,
+      "grad_norm": 1.9440633058547974,
+      "learning_rate": 8.475986451236623e-06,
+      "loss": 0.13878930807113649,
+      "mean_token_accuracy": 0.9542667560088329,
+      "num_tokens": 8098232.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3138233663493328,
+      "eval_loss": 0.8311719298362732,
+      "eval_mean_token_accuracy": 0.8295681362741449,
+      "eval_num_tokens": 8098232.0,
+      "eval_runtime": 98.401,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3320
+    },
+    {
+      "entropy": 0.19024005569517613,
+      "epoch": 8.048250904704464,
+      "grad_norm": 0.6860044598579407,
+      "learning_rate": 8.088677824972262e-06,
+      "loss": 0.104938805103302,
+      "mean_token_accuracy": 0.9669251158833504,
+      "num_tokens": 8145004.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.048250904704464,
+      "eval_entropy": 0.2925276593043563,
+      "eval_loss": 0.9170228838920593,
+      "eval_mean_token_accuracy": 0.8260884147681548,
+      "eval_num_tokens": 8145004.0,
+      "eval_runtime": 98.3738,
+      "eval_samples_per_second": 14.435,
+      "eval_steps_per_second": 1.809,
+      "step": 3340
+    },
+    {
+      "entropy": 0.17429662458598613,
+      "epoch": 8.096501809408926,
+      "grad_norm": 0.6853544116020203,
+      "learning_rate": 7.70930893970481e-06,
+      "loss": 0.09820220470428467,
+      "mean_token_accuracy": 0.9690332591533661,
+      "num_tokens": 8193186.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.096501809408926,
+      "eval_entropy": 0.29154844699280985,
+      "eval_loss": 0.9182955026626587,
+      "eval_mean_token_accuracy": 0.8260592245653774,
+      "eval_num_tokens": 8193186.0,
+      "eval_runtime": 98.5024,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3360
+    },
+    {
+      "entropy": 0.17998851463198662,
+      "epoch": 8.14475271411339,
+      "grad_norm": 0.6730113625526428,
+      "learning_rate": 7.337987152341293e-06,
+      "loss": 0.10509306192398071,
+      "mean_token_accuracy": 0.9663080736994744,
+      "num_tokens": 8241385.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.14475271411339,
+      "eval_entropy": 0.29224191346530165,
+      "eval_loss": 0.9142967462539673,
+      "eval_mean_token_accuracy": 0.825743991672323,
+      "eval_num_tokens": 8241385.0,
+      "eval_runtime": 98.3875,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3380
+    },
+    {
+      "entropy": 0.16425799671560526,
+      "epoch": 8.193003618817853,
+      "grad_norm": 0.8334925174713135,
+      "learning_rate": 6.974817542555239e-06,
+      "loss": 0.09438974857330322,
+      "mean_token_accuracy": 0.9704208940267562,
+      "num_tokens": 8294458.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.193003618817853,
+      "eval_entropy": 0.2905296404542548,
+      "eval_loss": 0.9226241707801819,
+      "eval_mean_token_accuracy": 0.8262435773785195,
+      "eval_num_tokens": 8294458.0,
+      "eval_runtime": 98.6148,
+      "eval_samples_per_second": 14.399,
+      "eval_steps_per_second": 1.805,
+      "step": 3400
+    },
+    {
+      "entropy": 0.16865960750728845,
+      "epoch": 8.241254523522317,
+      "grad_norm": 0.7445836663246155,
+      "learning_rate": 6.619902883050312e-06,
+      "loss": 0.09588083028793334,
+      "mean_token_accuracy": 0.9690534204244614,
+      "num_tokens": 8347663.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.241254523522317,
+      "eval_entropy": 0.28748046489578954,
+      "eval_loss": 0.9263729453086853,
+      "eval_mean_token_accuracy": 0.8265480908115258,
+      "eval_num_tokens": 8347663.0,
+      "eval_runtime": 98.5098,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3420
+    },
+    {
+      "entropy": 0.1850855764001608,
+      "epoch": 8.289505428226779,
+      "grad_norm": 0.8839195370674133,
+      "learning_rate": 6.273343610476938e-06,
+      "loss": 0.1067918062210083,
+      "mean_token_accuracy": 0.9668746247887612,
+      "num_tokens": 8391318.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.289505428226779,
+      "eval_entropy": 0.2879818446850509,
+      "eval_loss": 0.9265933036804199,
+      "eval_mean_token_accuracy": 0.826676292700714,
+      "eval_num_tokens": 8391318.0,
+      "eval_runtime": 98.491,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3440
+    },
+    {
+      "entropy": 0.1764491790905595,
+      "epoch": 8.337756332931242,
+      "grad_norm": 0.8340020775794983,
+      "learning_rate": 5.935237797009842e-06,
+      "loss": 0.10066965818405152,
+      "mean_token_accuracy": 0.9678841292858124,
+      "num_tokens": 8440740.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.337756332931242,
+      "eval_entropy": 0.28914196303721224,
+      "eval_loss": 0.9251464605331421,
+      "eval_mean_token_accuracy": 0.8263801664448855,
+      "eval_num_tokens": 8440740.0,
+      "eval_runtime": 98.366,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3460
+    },
+    {
+      "entropy": 0.17226074449717999,
+      "epoch": 8.386007237635706,
+      "grad_norm": 0.7149388194084167,
+      "learning_rate": 5.605681122594823e-06,
+      "loss": 0.10052882432937622,
+      "mean_token_accuracy": 0.9679374754428863,
+      "num_tokens": 8490767.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.386007237635706,
+      "eval_entropy": 0.28626472519689733,
+      "eval_loss": 0.9307431578636169,
+      "eval_mean_token_accuracy": 0.8269837619213576,
+      "eval_num_tokens": 8490767.0,
+      "eval_runtime": 98.3539,
+      "eval_samples_per_second": 14.438,
+      "eval_steps_per_second": 1.81,
+      "step": 3480
+    },
+    {
+      "entropy": 0.17341369036585091,
+      "epoch": 8.43425814234017,
+      "grad_norm": 0.6688793301582336,
+      "learning_rate": 5.28476684787248e-06,
+      "loss": 0.10054340362548828,
+      "mean_token_accuracy": 0.9675475805997849,
+      "num_tokens": 8538959.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.43425814234017,
+      "eval_entropy": 0.28786867212378575,
+      "eval_loss": 0.9303860664367676,
+      "eval_mean_token_accuracy": 0.8259982466697693,
+      "eval_num_tokens": 8538959.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3500
+    },
+    {
+      "entropy": 0.16994166951626538,
+      "epoch": 8.482509047044632,
+      "grad_norm": 0.7877247333526611,
+      "learning_rate": 4.972585787786484e-06,
+      "loss": 0.0993818998336792,
+      "mean_token_accuracy": 0.9688723921775818,
+      "num_tokens": 8587583.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.482509047044632,
+      "eval_entropy": 0.287084077600013,
+      "eval_loss": 0.9293763041496277,
+      "eval_mean_token_accuracy": 0.8268557224380836,
+      "eval_num_tokens": 8587583.0,
+      "eval_runtime": 98.5488,
+      "eval_samples_per_second": 14.409,
+      "eval_steps_per_second": 1.806,
+      "step": 3520
+    },
+    {
+      "entropy": 0.1698782866820693,
+      "epoch": 8.530759951749095,
+      "grad_norm": 0.6359167695045471,
+      "learning_rate": 4.66922628588412e-06,
+      "loss": 0.10011144876480102,
+      "mean_token_accuracy": 0.968739403784275,
+      "num_tokens": 8636960.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.530759951749095,
+      "eval_entropy": 0.2874377721481109,
+      "eval_loss": 0.9322494268417358,
+      "eval_mean_token_accuracy": 0.8263491124249576,
+      "eval_num_tokens": 8636960.0,
+      "eval_runtime": 98.3916,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3540
+    },
+    {
+      "entropy": 0.17174113132059574,
+      "epoch": 8.579010856453559,
+      "grad_norm": 0.7202840447425842,
+      "learning_rate": 4.374774189316017e-06,
+      "loss": 0.10309455394744874,
+      "mean_token_accuracy": 0.9690345630049706,
+      "num_tokens": 8685482.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.579010856453559,
+      "eval_entropy": 0.286078037673168,
+      "eval_loss": 0.9365394115447998,
+      "eval_mean_token_accuracy": 0.8265787784972888,
+      "eval_num_tokens": 8685482.0,
+      "eval_runtime": 98.5151,
+      "eval_samples_per_second": 14.414,
+      "eval_steps_per_second": 1.807,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1779118437319994,
+      "epoch": 8.627261761158021,
+      "grad_norm": 0.7755381464958191,
+      "learning_rate": 4.08931282454246e-06,
+      "loss": 0.10361287593841553,
+      "mean_token_accuracy": 0.9668692052364349,
+      "num_tokens": 8733512.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.627261761158021,
+      "eval_entropy": 0.2901240846749102,
+      "eval_loss": 0.9282606244087219,
+      "eval_mean_token_accuracy": 0.8257483565405513,
+      "eval_num_tokens": 8733512.0,
+      "eval_runtime": 98.3953,
+      "eval_samples_per_second": 14.432,
+      "eval_steps_per_second": 1.809,
+      "step": 3580
+    },
+    {
+      "entropy": 0.17399974688887596,
+      "epoch": 8.675512665862485,
+      "grad_norm": 0.6068329215049744,
+      "learning_rate": 3.8129229737530146e-06,
+      "loss": 0.10165332555770874,
+      "mean_token_accuracy": 0.9683542385697365,
+      "num_tokens": 8781726.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.675512665862485,
+      "eval_entropy": 0.28656756911384923,
+      "eval_loss": 0.93599933385849,
+      "eval_mean_token_accuracy": 0.8266928818788422,
+      "eval_num_tokens": 8781726.0,
+      "eval_runtime": 98.4991,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3600
+    },
+    {
+      "entropy": 0.17109180577099323,
+      "epoch": 8.723763570566948,
+      "grad_norm": 0.671509325504303,
+      "learning_rate": 3.5456828520060676e-06,
+      "loss": 0.10388398170471191,
+      "mean_token_accuracy": 0.9666652023792267,
+      "num_tokens": 8831964.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.723763570566948,
+      "eval_entropy": 0.28785753225007754,
+      "eval_loss": 0.9345396161079407,
+      "eval_mean_token_accuracy": 0.8258346751164855,
+      "eval_num_tokens": 8831964.0,
+      "eval_runtime": 98.5634,
+      "eval_samples_per_second": 14.407,
+      "eval_steps_per_second": 1.806,
+      "step": 3620
+    },
+    {
+      "entropy": 0.17080388870090246,
+      "epoch": 8.772014475271412,
+      "grad_norm": 0.9064943194389343,
+      "learning_rate": 3.2876680850949342e-06,
+      "loss": 0.09955486059188842,
+      "mean_token_accuracy": 0.9688052341341973,
+      "num_tokens": 8881274.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.772014475271412,
+      "eval_entropy": 0.2863340745313784,
+      "eval_loss": 0.9368643760681152,
+      "eval_mean_token_accuracy": 0.826385034604019,
+      "eval_num_tokens": 8881274.0,
+      "eval_runtime": 98.4956,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1739195568487048,
+      "epoch": 8.820265379975874,
+      "grad_norm": 0.645410418510437,
+      "learning_rate": 3.0389516881466266e-06,
+      "loss": 0.1038020372390747,
+      "mean_token_accuracy": 0.965588228404522,
+      "num_tokens": 8928303.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.820265379975874,
+      "eval_entropy": 0.2869644621282481,
+      "eval_loss": 0.9337670803070068,
+      "eval_mean_token_accuracy": 0.8266236021947325,
+      "eval_num_tokens": 8928303.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3660
+    },
+    {
+      "entropy": 0.16863423213362694,
+      "epoch": 8.868516284680338,
+      "grad_norm": 0.6110771298408508,
+      "learning_rate": 2.7996040449594445e-06,
+      "loss": 0.10157880783081055,
+      "mean_token_accuracy": 0.967547157406807,
+      "num_tokens": 8977325.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.868516284680338,
+      "eval_entropy": 0.28859044577968257,
+      "eval_loss": 0.9297890067100525,
+      "eval_mean_token_accuracy": 0.8264903181724335,
+      "eval_num_tokens": 8977325.0,
+      "eval_runtime": 98.5054,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1767999932169914,
+      "epoch": 8.916767189384801,
+      "grad_norm": 0.6543433666229248,
+      "learning_rate": 2.5696928880852117e-06,
+      "loss": 0.10324364900588989,
+      "mean_token_accuracy": 0.9673267215490341,
+      "num_tokens": 9026898.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.916767189384801,
+      "eval_entropy": 0.2885892426364877,
+      "eval_loss": 0.928764820098877,
+      "eval_mean_token_accuracy": 0.8265983676642514,
+      "eval_num_tokens": 9026898.0,
+      "eval_runtime": 98.4914,
+      "eval_samples_per_second": 14.418,
+      "eval_steps_per_second": 1.807,
+      "step": 3700
+    },
+    {
+      "entropy": 0.16943628452718257,
+      "epoch": 8.965018094089265,
+      "grad_norm": 0.8403868079185486,
+      "learning_rate": 2.3492832796617217e-06,
+      "loss": 0.09948049783706665,
+      "mean_token_accuracy": 0.967706835269928,
+      "num_tokens": 9077388.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.965018094089265,
+      "eval_entropy": 0.2884856275795551,
+      "eval_loss": 0.933276355266571,
+      "eval_mean_token_accuracy": 0.8261616119507993,
+      "eval_num_tokens": 9077388.0,
+      "eval_runtime": 98.6259,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3720
+    },
+    {
+      "entropy": 0.17359357613783616,
+      "epoch": 9.012062726176115,
+      "grad_norm": 0.37213996052742004,
+      "learning_rate": 2.138437593000977e-06,
+      "loss": 0.10089871883392335,
+      "mean_token_accuracy": 0.9696638874518566,
+      "num_tokens": 9122597.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.012062726176115,
+      "eval_entropy": 0.2875421148505104,
+      "eval_loss": 0.9357590079307556,
+      "eval_mean_token_accuracy": 0.8264994859025719,
+      "eval_num_tokens": 9122597.0,
+      "eval_runtime": 98.523,
+      "eval_samples_per_second": 14.413,
+      "eval_steps_per_second": 1.807,
+      "step": 3740
+    },
+    {
+      "entropy": 0.1642639346420765,
+      "epoch": 9.060313630880579,
+      "grad_norm": 0.5511618852615356,
+      "learning_rate": 1.937215494938237e-06,
+      "loss": 0.08524587750434875,
+      "mean_token_accuracy": 0.9754035726189614,
+      "num_tokens": 9171477.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.060313630880579,
+      "eval_entropy": 0.2820046850469675,
+      "eval_loss": 0.9588018655776978,
+      "eval_mean_token_accuracy": 0.8257443647036392,
+      "eval_num_tokens": 9171477.0,
+      "eval_runtime": 98.3992,
+      "eval_samples_per_second": 14.431,
+      "eval_steps_per_second": 1.809,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1630103051662445,
+      "epoch": 9.108564535585042,
+      "grad_norm": 0.5125291347503662,
+      "learning_rate": 1.7456739289470143e-06,
+      "loss": 0.0895907461643219,
+      "mean_token_accuracy": 0.973100857436657,
+      "num_tokens": 9217543.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.108564535585042,
+      "eval_entropy": 0.2786683516388529,
+      "eval_loss": 0.9743365049362183,
+      "eval_mean_token_accuracy": 0.8253507131940863,
+      "eval_num_tokens": 9217543.0,
+      "eval_runtime": 98.3882,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3780
+    },
+    {
+      "entropy": 0.15873534101992845,
+      "epoch": 9.156815440289506,
+      "grad_norm": 0.5697775483131409,
+      "learning_rate": 1.5638670990247788e-06,
+      "loss": 0.08552997708320617,
+      "mean_token_accuracy": 0.973672679066658,
+      "num_tokens": 9268724.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.156815440289506,
+      "eval_entropy": 0.2781428360704626,
+      "eval_loss": 0.9790021777153015,
+      "eval_mean_token_accuracy": 0.8251898780967413,
+      "eval_num_tokens": 9268724.0,
+      "eval_runtime": 98.5449,
+      "eval_samples_per_second": 14.41,
+      "eval_steps_per_second": 1.806,
+      "step": 3800
+    },
+    {
+      "entropy": 0.1606229657307267,
+      "epoch": 9.205066344993968,
+      "grad_norm": 0.5715742707252502,
+      "learning_rate": 1.3918464543538225e-06,
+      "loss": 0.08984670042991638,
+      "mean_token_accuracy": 0.9729854613542557,
+      "num_tokens": 9316805.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.205066344993968,
+      "eval_entropy": 0.2777416408899125,
+      "eval_loss": 0.9808777570724487,
+      "eval_mean_token_accuracy": 0.8252255926641185,
+      "eval_num_tokens": 9316805.0,
+      "eval_runtime": 98.627,
+      "eval_samples_per_second": 14.398,
+      "eval_steps_per_second": 1.805,
+      "step": 3820
+    },
+    {
+      "entropy": 0.15027269292622805,
+      "epoch": 9.253317249698432,
+      "grad_norm": 0.3359828293323517,
+      "learning_rate": 1.229660674741786e-06,
+      "loss": 0.08229721784591675,
+      "mean_token_accuracy": 0.9754723474383354,
+      "num_tokens": 9369940.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.253317249698432,
+      "eval_entropy": 0.2768587637483404,
+      "eval_loss": 0.9837455749511719,
+      "eval_mean_token_accuracy": 0.825493957889214,
+      "eval_num_tokens": 9369940.0,
+      "eval_runtime": 98.534,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 1.806,
+      "step": 3840
+    },
+    {
+      "entropy": 0.16767099779099226,
+      "epoch": 9.301568154402895,
+      "grad_norm": 0.5708875060081482,
+      "learning_rate": 1.077355656845822e-06,
+      "loss": 0.08888025283813476,
+      "mean_token_accuracy": 0.9715678468346596,
+      "num_tokens": 9416788.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.301568154402895,
+      "eval_entropy": 0.27612990283229377,
+      "eval_loss": 0.9873999357223511,
+      "eval_mean_token_accuracy": 0.8252144948150335,
+      "eval_num_tokens": 9416788.0,
+      "eval_runtime": 98.3889,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3860
+    },
+    {
+      "entropy": 0.1549418892711401,
+      "epoch": 9.349819059107359,
+      "grad_norm": 0.7601456046104431,
+      "learning_rate": 9.349745011844029e-07,
+      "loss": 0.08411069512367249,
+      "mean_token_accuracy": 0.9741103485226631,
+      "num_tokens": 9467862.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.349819059107359,
+      "eval_entropy": 0.2757877716689967,
+      "eval_loss": 0.9892058968544006,
+      "eval_mean_token_accuracy": 0.8250612017143978,
+      "eval_num_tokens": 9467862.0,
+      "eval_runtime": 98.3869,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 3880
+    },
+    {
+      "entropy": 0.16111517809331416,
+      "epoch": 9.398069963811821,
+      "grad_norm": 0.6774276494979858,
+      "learning_rate": 8.025574999403695e-07,
+      "loss": 0.08765592575073242,
+      "mean_token_accuracy": 0.9734174907207489,
+      "num_tokens": 9514611.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.398069963811821,
+      "eval_entropy": 0.275664389049739,
+      "eval_loss": 0.989257276058197,
+      "eval_mean_token_accuracy": 0.8252896914321385,
+      "eval_num_tokens": 9514611.0,
+      "eval_runtime": 98.4649,
+      "eval_samples_per_second": 14.421,
+      "eval_steps_per_second": 1.808,
+      "step": 3900
+    },
+    {
+      "entropy": 0.15701593235135078,
+      "epoch": 9.446320868516285,
+      "grad_norm": 0.6130298376083374,
+      "learning_rate": 6.801421255587377e-07,
+      "loss": 0.0853091835975647,
+      "mean_token_accuracy": 0.9740088582038879,
+      "num_tokens": 9563437.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.446320868516285,
+      "eval_entropy": 0.27559053881114787,
+      "eval_loss": 0.9901270866394043,
+      "eval_mean_token_accuracy": 0.8251507653948966,
+      "eval_num_tokens": 9563437.0,
+      "eval_runtime": 98.5285,
+      "eval_samples_per_second": 14.412,
+      "eval_steps_per_second": 1.807,
+      "step": 3920
+    },
+    {
+      "entropy": 0.15848954878747462,
+      "epoch": 9.494571773220748,
+      "grad_norm": 0.5928868651390076,
+      "learning_rate": 5.677630201424204e-07,
+      "loss": 0.08467000126838684,
+      "mean_token_accuracy": 0.9725634664297104,
+      "num_tokens": 9614034.0,
+      "step": 3940
+    },
+    {
+      "epoch": 9.494571773220748,
+      "eval_entropy": 0.2750965692521481,
+      "eval_loss": 0.9914345145225525,
+      "eval_mean_token_accuracy": 0.8252089954494091,
+      "eval_num_tokens": 9614034.0,
+      "eval_runtime": 98.5049,
+      "eval_samples_per_second": 14.416,
+      "eval_steps_per_second": 1.807,
+      "step": 3940
+    },
+    {
+      "entropy": 0.15644967313855887,
+      "epoch": 9.54282267792521,
+      "grad_norm": 0.5718610882759094,
+      "learning_rate": 4.6545198564893463e-07,
+      "loss": 0.08564762473106384,
+      "mean_token_accuracy": 0.9730244860053062,
+      "num_tokens": 9661266.0,
+      "step": 3960
+    },
+    {
+      "epoch": 9.54282267792521,
+      "eval_entropy": 0.27522430820076654,
+      "eval_loss": 0.9908939003944397,
+      "eval_mean_token_accuracy": 0.8251052994406625,
+      "eval_num_tokens": 9661266.0,
+      "eval_runtime": 98.3637,
+      "eval_samples_per_second": 14.436,
+      "eval_steps_per_second": 1.81,
+      "step": 3960
+    },
+    {
+      "entropy": 0.15196608882397414,
+      "epoch": 9.591073582629674,
+      "grad_norm": 0.752661406993866,
+      "learning_rate": 3.732379748908161e-07,
+      "loss": 0.08072127103805542,
+      "mean_token_accuracy": 0.9749975100159645,
+      "num_tokens": 9710458.0,
+      "step": 3980
+    },
+    {
+      "epoch": 9.591073582629674,
+      "eval_entropy": 0.2750371985890892,
+      "eval_loss": 0.9913897514343262,
+      "eval_mean_token_accuracy": 0.8248539731073915,
+      "eval_num_tokens": 9710458.0,
+      "eval_runtime": 98.4959,
+      "eval_samples_per_second": 14.417,
+      "eval_steps_per_second": 1.807,
+      "step": 3980
+    },
+    {
+      "entropy": 0.16616767533123494,
+      "epoch": 9.639324487334138,
+      "grad_norm": 0.7107775211334229,
+      "learning_rate": 2.911470833423274e-07,
+      "loss": 0.08988997340202332,
+      "mean_token_accuracy": 0.9714749112725258,
+      "num_tokens": 9755797.0,
+      "step": 4000
+    },
+    {
+      "epoch": 9.639324487334138,
+      "eval_entropy": 0.27523158457172053,
+      "eval_loss": 0.9913759827613831,
+      "eval_mean_token_accuracy": 0.8250464505693885,
+      "eval_num_tokens": 9755797.0,
+      "eval_runtime": 98.5067,
+      "eval_samples_per_second": 14.415,
+      "eval_steps_per_second": 1.807,
+      "step": 4000
+    },
+    {
+      "entropy": 0.15760798901319503,
+      "epoch": 9.687575392038601,
+      "grad_norm": 0.5563387870788574,
+      "learning_rate": 2.1920254175472834e-07,
+      "loss": 0.0877057135105133,
+      "mean_token_accuracy": 0.9734948605298996,
+      "num_tokens": 9805552.0,
+      "step": 4020
+    },
+    {
+      "epoch": 9.687575392038601,
+      "eval_entropy": 0.27528667726208655,
+      "eval_loss": 0.9914706349372864,
+      "eval_mean_token_accuracy": 0.8250808802883277,
+      "eval_num_tokens": 9805552.0,
+      "eval_runtime": 98.475,
+      "eval_samples_per_second": 14.42,
+      "eval_steps_per_second": 1.808,
+      "step": 4020
+    },
+    {
+      "entropy": 0.161653464846313,
+      "epoch": 9.735826296743063,
+      "grad_norm": 0.5896415114402771,
+      "learning_rate": 1.5742470958227596e-07,
+      "loss": 0.08775997757911683,
+      "mean_token_accuracy": 0.9718310445547104,
+      "num_tokens": 9852843.0,
+      "step": 4040
+    },
+    {
+      "epoch": 9.735826296743063,
+      "eval_entropy": 0.27531274182073184,
+      "eval_loss": 0.9915319085121155,
+      "eval_mean_token_accuracy": 0.825061980258213,
+      "eval_num_tokens": 9852843.0,
+      "eval_runtime": 98.3873,
+      "eval_samples_per_second": 14.433,
+      "eval_steps_per_second": 1.809,
+      "step": 4040
+    },
+    {
+      "entropy": 0.16010624952614308,
+      "epoch": 9.784077201447527,
+      "grad_norm": 0.6022347211837769,
+      "learning_rate": 1.058310692207063e-07,
+      "loss": 0.08761411905288696,
+      "mean_token_accuracy": 0.9729624018073082,
+      "num_tokens": 9900151.0,
+      "step": 4060
+    },
+    {
+      "epoch": 9.784077201447527,
+      "eval_entropy": 0.27524732446737504,
+      "eval_loss": 0.991264820098877,
+      "eval_mean_token_accuracy": 0.8251047007153544,
+      "eval_num_tokens": 9900151.0,
+      "eval_runtime": 98.4834,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 4060
+    },
+    {
+      "entropy": 0.159021489135921,
+      "epoch": 9.83232810615199,
+      "grad_norm": 0.48451653122901917,
+      "learning_rate": 6.443622105994901e-08,
+      "loss": 0.08291574716567993,
+      "mean_token_accuracy": 0.9726693674921989,
+      "num_tokens": 9952926.0,
+      "step": 4080
+    },
+    {
+      "epoch": 9.83232810615199,
+      "eval_entropy": 0.27521907144717955,
+      "eval_loss": 0.9918096661567688,
+      "eval_mean_token_accuracy": 0.8250070894032382,
+      "eval_num_tokens": 9952926.0,
+      "eval_runtime": 98.3568,
+      "eval_samples_per_second": 14.437,
+      "eval_steps_per_second": 1.81,
+      "step": 4080
+    },
+    {
+      "entropy": 0.15175675600767136,
+      "epoch": 9.880579010856454,
+      "grad_norm": 0.5009680390357971,
+      "learning_rate": 3.325187935238231e-08,
+      "loss": 0.08214685916900635,
+      "mean_token_accuracy": 0.9752551406621933,
+      "num_tokens": 10003720.0,
+      "step": 4100
+    },
+    {
+      "epoch": 9.880579010856454,
+      "eval_entropy": 0.27522189488236826,
+      "eval_loss": 0.9918575882911682,
+      "eval_mean_token_accuracy": 0.8249149091458052,
+      "eval_num_tokens": 10003720.0,
+      "eval_runtime": 98.4802,
+      "eval_samples_per_second": 14.419,
+      "eval_steps_per_second": 1.807,
+      "step": 4100
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.074449495401472e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..1976b0e253c319a98984758c3212495e5aabf86f
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/trainer_state.json
@@ -0,0 +1,580 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.293897882938979,
+  "eval_steps": 20,
+  "global_step": 520,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.60897146535424e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..bcdadfda53e83a7371f6b73c85a20aa842817590
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/trainer_state.json
@@ -0,0 +1,601 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.3437110834371109,
+  "eval_steps": 20,
+  "global_step": 540,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.7055833422026752e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..bb4931f1a00afa706e8f3d1823de6b9ffcef1b18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/trainer_state.json
@@ -0,0 +1,622 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.3935242839352429,
+  "eval_steps": 20,
+  "global_step": 560,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.8063236937193472e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..661a81dce91f7ee34a59cb8f9e3d745ee8df29c9
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/trainer_state.json
@@ -0,0 +1,643 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.4433374844333748,
+  "eval_steps": 20,
+  "global_step": 580,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.902538735734579e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d7e6bbab5005c4a594b2a40373a75f38b31af574
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/trainer_state.json
@@ -0,0 +1,97 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.149439601494396,
+  "eval_steps": 20,
+  "global_step": 60,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.826160213868544e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..339ed894802cf0d818f707ed19f9c8242683497c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/trainer_state.json
@@ -0,0 +1,664 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.4931506849315068,
+  "eval_steps": 20,
+  "global_step": 600,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.002763898169549e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e6bf9cedb7d45a7e8935a9ec3eac41156a027854
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/trainer_state.json
@@ -0,0 +1,685 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.5429638854296388,
+  "eval_steps": 20,
+  "global_step": 620,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    },
+    {
+      "entropy": 0.49362134337425234,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.26279115676879883,
+      "learning_rate": 9.819570324736658e-05,
+      "loss": 0.45868754386901855,
+      "mean_token_accuracy": 0.8646741047501564,
+      "num_tokens": 1511476.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.4934931300407232,
+      "eval_loss": 0.47820955514907837,
+      "eval_mean_token_accuracy": 0.8611027102830798,
+      "eval_num_tokens": 1511476.0,
+      "eval_runtime": 96.865,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 620
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.115249168346112e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..88f74b3b8ea855140516b88150e02605421d3c1f
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/trainer_state.json
@@ -0,0 +1,706 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.592777085927771,
+  "eval_steps": 20,
+  "global_step": 640,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    },
+    {
+      "entropy": 0.49362134337425234,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.26279115676879883,
+      "learning_rate": 9.819570324736658e-05,
+      "loss": 0.45868754386901855,
+      "mean_token_accuracy": 0.8646741047501564,
+      "num_tokens": 1511476.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.4934931300407232,
+      "eval_loss": 0.47820955514907837,
+      "eval_mean_token_accuracy": 0.8611027102830798,
+      "eval_num_tokens": 1511476.0,
+      "eval_runtime": 96.865,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 620
+    },
+    {
+      "entropy": 0.498298654705286,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.2743106484413147,
+      "learning_rate": 9.802723416796326e-05,
+      "loss": 0.4625910758972168,
+      "mean_token_accuracy": 0.8644399270415306,
+      "num_tokens": 1560069.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.49935419944136644,
+      "eval_loss": 0.4788094460964203,
+      "eval_mean_token_accuracy": 0.8595240989396739,
+      "eval_num_tokens": 1560069.0,
+      "eval_runtime": 96.8858,
+      "eval_samples_per_second": 14.202,
+      "eval_steps_per_second": 1.775,
+      "step": 640
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.211547754536755e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..48cd9b9c58ef7724d2a4bb329d4ab8c67f888ce6
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/trainer_state.json
@@ -0,0 +1,727 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.6425902864259028,
+  "eval_steps": 20,
+  "global_step": 660,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    },
+    {
+      "entropy": 0.49362134337425234,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.26279115676879883,
+      "learning_rate": 9.819570324736658e-05,
+      "loss": 0.45868754386901855,
+      "mean_token_accuracy": 0.8646741047501564,
+      "num_tokens": 1511476.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.4934931300407232,
+      "eval_loss": 0.47820955514907837,
+      "eval_mean_token_accuracy": 0.8611027102830798,
+      "eval_num_tokens": 1511476.0,
+      "eval_runtime": 96.865,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 620
+    },
+    {
+      "entropy": 0.498298654705286,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.2743106484413147,
+      "learning_rate": 9.802723416796326e-05,
+      "loss": 0.4625910758972168,
+      "mean_token_accuracy": 0.8644399270415306,
+      "num_tokens": 1560069.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.49935419944136644,
+      "eval_loss": 0.4788094460964203,
+      "eval_mean_token_accuracy": 0.8595240989396739,
+      "eval_num_tokens": 1560069.0,
+      "eval_runtime": 96.8858,
+      "eval_samples_per_second": 14.202,
+      "eval_steps_per_second": 1.775,
+      "step": 640
+    },
+    {
+      "entropy": 0.5169856667518615,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.2438548356294632,
+      "learning_rate": 9.784414086634099e-05,
+      "loss": 0.4727236270904541,
+      "mean_token_accuracy": 0.8596569061279297,
+      "num_tokens": 1603766.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.48505116392706715,
+      "eval_loss": 0.4763250946998596,
+      "eval_mean_token_accuracy": 0.8609154699153678,
+      "eval_num_tokens": 1603766.0,
+      "eval_runtime": 96.8748,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 660
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.3018381221124096e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..63fac734dc4821a17222d0b932e63ad987350237
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/trainer_state.json
@@ -0,0 +1,748 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.692403486924035,
+  "eval_steps": 20,
+  "global_step": 680,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    },
+    {
+      "entropy": 0.49362134337425234,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.26279115676879883,
+      "learning_rate": 9.819570324736658e-05,
+      "loss": 0.45868754386901855,
+      "mean_token_accuracy": 0.8646741047501564,
+      "num_tokens": 1511476.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.4934931300407232,
+      "eval_loss": 0.47820955514907837,
+      "eval_mean_token_accuracy": 0.8611027102830798,
+      "eval_num_tokens": 1511476.0,
+      "eval_runtime": 96.865,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 620
+    },
+    {
+      "entropy": 0.498298654705286,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.2743106484413147,
+      "learning_rate": 9.802723416796326e-05,
+      "loss": 0.4625910758972168,
+      "mean_token_accuracy": 0.8644399270415306,
+      "num_tokens": 1560069.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.49935419944136644,
+      "eval_loss": 0.4788094460964203,
+      "eval_mean_token_accuracy": 0.8595240989396739,
+      "eval_num_tokens": 1560069.0,
+      "eval_runtime": 96.8858,
+      "eval_samples_per_second": 14.202,
+      "eval_steps_per_second": 1.775,
+      "step": 640
+    },
+    {
+      "entropy": 0.5169856667518615,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.2438548356294632,
+      "learning_rate": 9.784414086634099e-05,
+      "loss": 0.4727236270904541,
+      "mean_token_accuracy": 0.8596569061279297,
+      "num_tokens": 1603766.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.48505116392706715,
+      "eval_loss": 0.4763250946998596,
+      "eval_mean_token_accuracy": 0.8609154699153678,
+      "eval_num_tokens": 1603766.0,
+      "eval_runtime": 96.8748,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 660
+    },
+    {
+      "entropy": 0.48837797343730927,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.2245505303144455,
+      "learning_rate": 9.764647856094226e-05,
+      "loss": 0.45554113388061523,
+      "mean_token_accuracy": 0.8651937782764435,
+      "num_tokens": 1654672.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.47867353970921317,
+      "eval_loss": 0.47178956866264343,
+      "eval_mean_token_accuracy": 0.8621448333873305,
+      "eval_num_tokens": 1654672.0,
+      "eval_runtime": 96.8536,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 680
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.4096658045097984e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..30c90d4f1d42197d830c036e9d37d2854c48de52
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/trainer_state.json
@@ -0,0 +1,769 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.7422166874221667,
+  "eval_steps": 20,
+  "global_step": 700,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    },
+    {
+      "entropy": 0.49362134337425234,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.26279115676879883,
+      "learning_rate": 9.819570324736658e-05,
+      "loss": 0.45868754386901855,
+      "mean_token_accuracy": 0.8646741047501564,
+      "num_tokens": 1511476.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.4934931300407232,
+      "eval_loss": 0.47820955514907837,
+      "eval_mean_token_accuracy": 0.8611027102830798,
+      "eval_num_tokens": 1511476.0,
+      "eval_runtime": 96.865,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 620
+    },
+    {
+      "entropy": 0.498298654705286,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.2743106484413147,
+      "learning_rate": 9.802723416796326e-05,
+      "loss": 0.4625910758972168,
+      "mean_token_accuracy": 0.8644399270415306,
+      "num_tokens": 1560069.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.49935419944136644,
+      "eval_loss": 0.4788094460964203,
+      "eval_mean_token_accuracy": 0.8595240989396739,
+      "eval_num_tokens": 1560069.0,
+      "eval_runtime": 96.8858,
+      "eval_samples_per_second": 14.202,
+      "eval_steps_per_second": 1.775,
+      "step": 640
+    },
+    {
+      "entropy": 0.5169856667518615,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.2438548356294632,
+      "learning_rate": 9.784414086634099e-05,
+      "loss": 0.4727236270904541,
+      "mean_token_accuracy": 0.8596569061279297,
+      "num_tokens": 1603766.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.48505116392706715,
+      "eval_loss": 0.4763250946998596,
+      "eval_mean_token_accuracy": 0.8609154699153678,
+      "eval_num_tokens": 1603766.0,
+      "eval_runtime": 96.8748,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 660
+    },
+    {
+      "entropy": 0.48837797343730927,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.2245505303144455,
+      "learning_rate": 9.764647856094226e-05,
+      "loss": 0.45554113388061523,
+      "mean_token_accuracy": 0.8651937782764435,
+      "num_tokens": 1654672.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.47867353970921317,
+      "eval_loss": 0.47178956866264343,
+      "eval_mean_token_accuracy": 0.8621448333873305,
+      "eval_num_tokens": 1654672.0,
+      "eval_runtime": 96.8536,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 680
+    },
+    {
+      "entropy": 0.4763235367834568,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.24530072510242462,
+      "learning_rate": 9.743430686402244e-05,
+      "loss": 0.44700350761413576,
+      "mean_token_accuracy": 0.8670215323567391,
+      "num_tokens": 1706860.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5032368693933931,
+      "eval_loss": 0.466962605714798,
+      "eval_mean_token_accuracy": 0.8626160392927569,
+      "eval_num_tokens": 1706860.0,
+      "eval_runtime": 96.8466,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 700
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.5114991923214336e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..83778a0b47872d1d56a689ceb8c043c6fc402b7d
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/trainer_state.json
@@ -0,0 +1,790 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.792029887920299,
+  "eval_steps": 20,
+  "global_step": 720,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    },
+    {
+      "entropy": 0.49362134337425234,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.26279115676879883,
+      "learning_rate": 9.819570324736658e-05,
+      "loss": 0.45868754386901855,
+      "mean_token_accuracy": 0.8646741047501564,
+      "num_tokens": 1511476.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.4934931300407232,
+      "eval_loss": 0.47820955514907837,
+      "eval_mean_token_accuracy": 0.8611027102830798,
+      "eval_num_tokens": 1511476.0,
+      "eval_runtime": 96.865,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 620
+    },
+    {
+      "entropy": 0.498298654705286,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.2743106484413147,
+      "learning_rate": 9.802723416796326e-05,
+      "loss": 0.4625910758972168,
+      "mean_token_accuracy": 0.8644399270415306,
+      "num_tokens": 1560069.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.49935419944136644,
+      "eval_loss": 0.4788094460964203,
+      "eval_mean_token_accuracy": 0.8595240989396739,
+      "eval_num_tokens": 1560069.0,
+      "eval_runtime": 96.8858,
+      "eval_samples_per_second": 14.202,
+      "eval_steps_per_second": 1.775,
+      "step": 640
+    },
+    {
+      "entropy": 0.5169856667518615,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.2438548356294632,
+      "learning_rate": 9.784414086634099e-05,
+      "loss": 0.4727236270904541,
+      "mean_token_accuracy": 0.8596569061279297,
+      "num_tokens": 1603766.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.48505116392706715,
+      "eval_loss": 0.4763250946998596,
+      "eval_mean_token_accuracy": 0.8609154699153678,
+      "eval_num_tokens": 1603766.0,
+      "eval_runtime": 96.8748,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 660
+    },
+    {
+      "entropy": 0.48837797343730927,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.2245505303144455,
+      "learning_rate": 9.764647856094226e-05,
+      "loss": 0.45554113388061523,
+      "mean_token_accuracy": 0.8651937782764435,
+      "num_tokens": 1654672.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.47867353970921317,
+      "eval_loss": 0.47178956866264343,
+      "eval_mean_token_accuracy": 0.8621448333873305,
+      "eval_num_tokens": 1654672.0,
+      "eval_runtime": 96.8536,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 680
+    },
+    {
+      "entropy": 0.4763235367834568,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.24530072510242462,
+      "learning_rate": 9.743430686402244e-05,
+      "loss": 0.44700350761413576,
+      "mean_token_accuracy": 0.8670215323567391,
+      "num_tokens": 1706860.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5032368693933931,
+      "eval_loss": 0.466962605714798,
+      "eval_mean_token_accuracy": 0.8626160392927569,
+      "eval_num_tokens": 1706860.0,
+      "eval_runtime": 96.8466,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 700
+    },
+    {
+      "entropy": 0.48205163702368736,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.335552453994751,
+      "learning_rate": 9.720768976367141e-05,
+      "loss": 0.4490774631500244,
+      "mean_token_accuracy": 0.8643530890345573,
+      "num_tokens": 1754431.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.4913890836543815,
+      "eval_loss": 0.46917298436164856,
+      "eval_mean_token_accuracy": 0.862456685235334,
+      "eval_num_tokens": 1754431.0,
+      "eval_runtime": 96.875,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 720
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.604922466475008e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..451c5c968821532549e2ea0de5ca372801092130
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/trainer_state.json
@@ -0,0 +1,811 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.841843088418431,
+  "eval_steps": 20,
+  "global_step": 740,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    },
+    {
+      "entropy": 0.49362134337425234,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.26279115676879883,
+      "learning_rate": 9.819570324736658e-05,
+      "loss": 0.45868754386901855,
+      "mean_token_accuracy": 0.8646741047501564,
+      "num_tokens": 1511476.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.4934931300407232,
+      "eval_loss": 0.47820955514907837,
+      "eval_mean_token_accuracy": 0.8611027102830798,
+      "eval_num_tokens": 1511476.0,
+      "eval_runtime": 96.865,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 620
+    },
+    {
+      "entropy": 0.498298654705286,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.2743106484413147,
+      "learning_rate": 9.802723416796326e-05,
+      "loss": 0.4625910758972168,
+      "mean_token_accuracy": 0.8644399270415306,
+      "num_tokens": 1560069.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.49935419944136644,
+      "eval_loss": 0.4788094460964203,
+      "eval_mean_token_accuracy": 0.8595240989396739,
+      "eval_num_tokens": 1560069.0,
+      "eval_runtime": 96.8858,
+      "eval_samples_per_second": 14.202,
+      "eval_steps_per_second": 1.775,
+      "step": 640
+    },
+    {
+      "entropy": 0.5169856667518615,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.2438548356294632,
+      "learning_rate": 9.784414086634099e-05,
+      "loss": 0.4727236270904541,
+      "mean_token_accuracy": 0.8596569061279297,
+      "num_tokens": 1603766.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.48505116392706715,
+      "eval_loss": 0.4763250946998596,
+      "eval_mean_token_accuracy": 0.8609154699153678,
+      "eval_num_tokens": 1603766.0,
+      "eval_runtime": 96.8748,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 660
+    },
+    {
+      "entropy": 0.48837797343730927,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.2245505303144455,
+      "learning_rate": 9.764647856094226e-05,
+      "loss": 0.45554113388061523,
+      "mean_token_accuracy": 0.8651937782764435,
+      "num_tokens": 1654672.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.47867353970921317,
+      "eval_loss": 0.47178956866264343,
+      "eval_mean_token_accuracy": 0.8621448333873305,
+      "eval_num_tokens": 1654672.0,
+      "eval_runtime": 96.8536,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 680
+    },
+    {
+      "entropy": 0.4763235367834568,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.24530072510242462,
+      "learning_rate": 9.743430686402244e-05,
+      "loss": 0.44700350761413576,
+      "mean_token_accuracy": 0.8670215323567391,
+      "num_tokens": 1706860.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5032368693933931,
+      "eval_loss": 0.466962605714798,
+      "eval_mean_token_accuracy": 0.8626160392927569,
+      "eval_num_tokens": 1706860.0,
+      "eval_runtime": 96.8466,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 700
+    },
+    {
+      "entropy": 0.48205163702368736,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.335552453994751,
+      "learning_rate": 9.720768976367141e-05,
+      "loss": 0.4490774631500244,
+      "mean_token_accuracy": 0.8643530890345573,
+      "num_tokens": 1754431.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.4913890836543815,
+      "eval_loss": 0.46917298436164856,
+      "eval_mean_token_accuracy": 0.862456685235334,
+      "eval_num_tokens": 1754431.0,
+      "eval_runtime": 96.875,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 720
+    },
+    {
+      "entropy": 0.48629910722374914,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.2625334560871124,
+      "learning_rate": 9.696669560451573e-05,
+      "loss": 0.4475870132446289,
+      "mean_token_accuracy": 0.8631389886140823,
+      "num_tokens": 1804368.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.4911207033451213,
+      "eval_loss": 0.46702805161476135,
+      "eval_mean_token_accuracy": 0.8621654281782549,
+      "eval_num_tokens": 1804368.0,
+      "eval_runtime": 96.8685,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 740
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.706616613994291e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b1346dd991ce269bfba9a533cc150f094fa514fb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/trainer_state.json
@@ -0,0 +1,832 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.891656288916563,
+  "eval_steps": 20,
+  "global_step": 760,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    },
+    {
+      "entropy": 0.49362134337425234,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.26279115676879883,
+      "learning_rate": 9.819570324736658e-05,
+      "loss": 0.45868754386901855,
+      "mean_token_accuracy": 0.8646741047501564,
+      "num_tokens": 1511476.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.4934931300407232,
+      "eval_loss": 0.47820955514907837,
+      "eval_mean_token_accuracy": 0.8611027102830798,
+      "eval_num_tokens": 1511476.0,
+      "eval_runtime": 96.865,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 620
+    },
+    {
+      "entropy": 0.498298654705286,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.2743106484413147,
+      "learning_rate": 9.802723416796326e-05,
+      "loss": 0.4625910758972168,
+      "mean_token_accuracy": 0.8644399270415306,
+      "num_tokens": 1560069.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.49935419944136644,
+      "eval_loss": 0.4788094460964203,
+      "eval_mean_token_accuracy": 0.8595240989396739,
+      "eval_num_tokens": 1560069.0,
+      "eval_runtime": 96.8858,
+      "eval_samples_per_second": 14.202,
+      "eval_steps_per_second": 1.775,
+      "step": 640
+    },
+    {
+      "entropy": 0.5169856667518615,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.2438548356294632,
+      "learning_rate": 9.784414086634099e-05,
+      "loss": 0.4727236270904541,
+      "mean_token_accuracy": 0.8596569061279297,
+      "num_tokens": 1603766.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.48505116392706715,
+      "eval_loss": 0.4763250946998596,
+      "eval_mean_token_accuracy": 0.8609154699153678,
+      "eval_num_tokens": 1603766.0,
+      "eval_runtime": 96.8748,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 660
+    },
+    {
+      "entropy": 0.48837797343730927,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.2245505303144455,
+      "learning_rate": 9.764647856094226e-05,
+      "loss": 0.45554113388061523,
+      "mean_token_accuracy": 0.8651937782764435,
+      "num_tokens": 1654672.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.47867353970921317,
+      "eval_loss": 0.47178956866264343,
+      "eval_mean_token_accuracy": 0.8621448333873305,
+      "eval_num_tokens": 1654672.0,
+      "eval_runtime": 96.8536,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 680
+    },
+    {
+      "entropy": 0.4763235367834568,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.24530072510242462,
+      "learning_rate": 9.743430686402244e-05,
+      "loss": 0.44700350761413576,
+      "mean_token_accuracy": 0.8670215323567391,
+      "num_tokens": 1706860.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5032368693933931,
+      "eval_loss": 0.466962605714798,
+      "eval_mean_token_accuracy": 0.8626160392927569,
+      "eval_num_tokens": 1706860.0,
+      "eval_runtime": 96.8466,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 700
+    },
+    {
+      "entropy": 0.48205163702368736,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.335552453994751,
+      "learning_rate": 9.720768976367141e-05,
+      "loss": 0.4490774631500244,
+      "mean_token_accuracy": 0.8643530890345573,
+      "num_tokens": 1754431.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.4913890836543815,
+      "eval_loss": 0.46917298436164856,
+      "eval_mean_token_accuracy": 0.862456685235334,
+      "eval_num_tokens": 1754431.0,
+      "eval_runtime": 96.875,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 720
+    },
+    {
+      "entropy": 0.48629910722374914,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.2625334560871124,
+      "learning_rate": 9.696669560451573e-05,
+      "loss": 0.4475870132446289,
+      "mean_token_accuracy": 0.8631389886140823,
+      "num_tokens": 1804368.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.4911207033451213,
+      "eval_loss": 0.46702805161476135,
+      "eval_mean_token_accuracy": 0.8621654281782549,
+      "eval_num_tokens": 1804368.0,
+      "eval_runtime": 96.8685,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 740
+    },
+    {
+      "entropy": 0.475378530472517,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.2406812161207199,
+      "learning_rate": 9.671139706710683e-05,
+      "loss": 0.4375624179840088,
+      "mean_token_accuracy": 0.8660449728369712,
+      "num_tokens": 1850859.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.48983157044926356,
+      "eval_loss": 0.4654311239719391,
+      "eval_mean_token_accuracy": 0.8626760878535205,
+      "eval_num_tokens": 1850859.0,
+      "eval_runtime": 96.8581,
+      "eval_samples_per_second": 14.206,
+      "eval_steps_per_second": 1.776,
+      "step": 760
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.8009797601212416e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ea16bff19c64063a7368c9947549037d6c18a723
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/trainer_state.json
@@ -0,0 +1,853 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.9414694894146949,
+  "eval_steps": 20,
+  "global_step": 780,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    },
+    {
+      "entropy": 0.49362134337425234,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.26279115676879883,
+      "learning_rate": 9.819570324736658e-05,
+      "loss": 0.45868754386901855,
+      "mean_token_accuracy": 0.8646741047501564,
+      "num_tokens": 1511476.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.4934931300407232,
+      "eval_loss": 0.47820955514907837,
+      "eval_mean_token_accuracy": 0.8611027102830798,
+      "eval_num_tokens": 1511476.0,
+      "eval_runtime": 96.865,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 620
+    },
+    {
+      "entropy": 0.498298654705286,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.2743106484413147,
+      "learning_rate": 9.802723416796326e-05,
+      "loss": 0.4625910758972168,
+      "mean_token_accuracy": 0.8644399270415306,
+      "num_tokens": 1560069.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.49935419944136644,
+      "eval_loss": 0.4788094460964203,
+      "eval_mean_token_accuracy": 0.8595240989396739,
+      "eval_num_tokens": 1560069.0,
+      "eval_runtime": 96.8858,
+      "eval_samples_per_second": 14.202,
+      "eval_steps_per_second": 1.775,
+      "step": 640
+    },
+    {
+      "entropy": 0.5169856667518615,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.2438548356294632,
+      "learning_rate": 9.784414086634099e-05,
+      "loss": 0.4727236270904541,
+      "mean_token_accuracy": 0.8596569061279297,
+      "num_tokens": 1603766.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.48505116392706715,
+      "eval_loss": 0.4763250946998596,
+      "eval_mean_token_accuracy": 0.8609154699153678,
+      "eval_num_tokens": 1603766.0,
+      "eval_runtime": 96.8748,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 660
+    },
+    {
+      "entropy": 0.48837797343730927,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.2245505303144455,
+      "learning_rate": 9.764647856094226e-05,
+      "loss": 0.45554113388061523,
+      "mean_token_accuracy": 0.8651937782764435,
+      "num_tokens": 1654672.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.47867353970921317,
+      "eval_loss": 0.47178956866264343,
+      "eval_mean_token_accuracy": 0.8621448333873305,
+      "eval_num_tokens": 1654672.0,
+      "eval_runtime": 96.8536,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 680
+    },
+    {
+      "entropy": 0.4763235367834568,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.24530072510242462,
+      "learning_rate": 9.743430686402244e-05,
+      "loss": 0.44700350761413576,
+      "mean_token_accuracy": 0.8670215323567391,
+      "num_tokens": 1706860.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5032368693933931,
+      "eval_loss": 0.466962605714798,
+      "eval_mean_token_accuracy": 0.8626160392927569,
+      "eval_num_tokens": 1706860.0,
+      "eval_runtime": 96.8466,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 700
+    },
+    {
+      "entropy": 0.48205163702368736,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.335552453994751,
+      "learning_rate": 9.720768976367141e-05,
+      "loss": 0.4490774631500244,
+      "mean_token_accuracy": 0.8643530890345573,
+      "num_tokens": 1754431.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.4913890836543815,
+      "eval_loss": 0.46917298436164856,
+      "eval_mean_token_accuracy": 0.862456685235334,
+      "eval_num_tokens": 1754431.0,
+      "eval_runtime": 96.875,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 720
+    },
+    {
+      "entropy": 0.48629910722374914,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.2625334560871124,
+      "learning_rate": 9.696669560451573e-05,
+      "loss": 0.4475870132446289,
+      "mean_token_accuracy": 0.8631389886140823,
+      "num_tokens": 1804368.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.4911207033451213,
+      "eval_loss": 0.46702805161476135,
+      "eval_mean_token_accuracy": 0.8621654281782549,
+      "eval_num_tokens": 1804368.0,
+      "eval_runtime": 96.8685,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 740
+    },
+    {
+      "entropy": 0.475378530472517,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.2406812161207199,
+      "learning_rate": 9.671139706710683e-05,
+      "loss": 0.4375624179840088,
+      "mean_token_accuracy": 0.8660449728369712,
+      "num_tokens": 1850859.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.48983157044926356,
+      "eval_loss": 0.4654311239719391,
+      "eval_mean_token_accuracy": 0.8626760878535205,
+      "eval_num_tokens": 1850859.0,
+      "eval_runtime": 96.8581,
+      "eval_samples_per_second": 14.206,
+      "eval_steps_per_second": 1.776,
+      "step": 760
+    },
+    {
+      "entropy": 0.4725016601383686,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.2215867042541504,
+      "learning_rate": 9.644187114600152e-05,
+      "loss": 0.44096899032592773,
+      "mean_token_accuracy": 0.8674400866031646,
+      "num_tokens": 1901155.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.4720242848229963,
+      "eval_loss": 0.46539217233657837,
+      "eval_mean_token_accuracy": 0.8630018580791562,
+      "eval_num_tokens": 1901155.0,
+      "eval_runtime": 96.8222,
+      "eval_samples_per_second": 14.212,
+      "eval_steps_per_second": 1.776,
+      "step": 780
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.89920682436096e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c3bc856c8ed764aeb28438bacd87b2181f57f5b6
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/trainer_state.json
@@ -0,0 +1,118 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.199252801992528,
+  "eval_steps": 20,
+  "global_step": 80,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.9143927187456e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..9fdd7a211663a29005c22ae08ccd56a8d3aa4f22
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/trainer_state.json
@@ -0,0 +1,874 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.9912826899128269,
+  "eval_steps": 20,
+  "global_step": 800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    },
+    {
+      "entropy": 0.49362134337425234,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.26279115676879883,
+      "learning_rate": 9.819570324736658e-05,
+      "loss": 0.45868754386901855,
+      "mean_token_accuracy": 0.8646741047501564,
+      "num_tokens": 1511476.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.4934931300407232,
+      "eval_loss": 0.47820955514907837,
+      "eval_mean_token_accuracy": 0.8611027102830798,
+      "eval_num_tokens": 1511476.0,
+      "eval_runtime": 96.865,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 620
+    },
+    {
+      "entropy": 0.498298654705286,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.2743106484413147,
+      "learning_rate": 9.802723416796326e-05,
+      "loss": 0.4625910758972168,
+      "mean_token_accuracy": 0.8644399270415306,
+      "num_tokens": 1560069.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.49935419944136644,
+      "eval_loss": 0.4788094460964203,
+      "eval_mean_token_accuracy": 0.8595240989396739,
+      "eval_num_tokens": 1560069.0,
+      "eval_runtime": 96.8858,
+      "eval_samples_per_second": 14.202,
+      "eval_steps_per_second": 1.775,
+      "step": 640
+    },
+    {
+      "entropy": 0.5169856667518615,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.2438548356294632,
+      "learning_rate": 9.784414086634099e-05,
+      "loss": 0.4727236270904541,
+      "mean_token_accuracy": 0.8596569061279297,
+      "num_tokens": 1603766.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.48505116392706715,
+      "eval_loss": 0.4763250946998596,
+      "eval_mean_token_accuracy": 0.8609154699153678,
+      "eval_num_tokens": 1603766.0,
+      "eval_runtime": 96.8748,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 660
+    },
+    {
+      "entropy": 0.48837797343730927,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.2245505303144455,
+      "learning_rate": 9.764647856094226e-05,
+      "loss": 0.45554113388061523,
+      "mean_token_accuracy": 0.8651937782764435,
+      "num_tokens": 1654672.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.47867353970921317,
+      "eval_loss": 0.47178956866264343,
+      "eval_mean_token_accuracy": 0.8621448333873305,
+      "eval_num_tokens": 1654672.0,
+      "eval_runtime": 96.8536,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 680
+    },
+    {
+      "entropy": 0.4763235367834568,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.24530072510242462,
+      "learning_rate": 9.743430686402244e-05,
+      "loss": 0.44700350761413576,
+      "mean_token_accuracy": 0.8670215323567391,
+      "num_tokens": 1706860.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5032368693933931,
+      "eval_loss": 0.466962605714798,
+      "eval_mean_token_accuracy": 0.8626160392927569,
+      "eval_num_tokens": 1706860.0,
+      "eval_runtime": 96.8466,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 700
+    },
+    {
+      "entropy": 0.48205163702368736,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.335552453994751,
+      "learning_rate": 9.720768976367141e-05,
+      "loss": 0.4490774631500244,
+      "mean_token_accuracy": 0.8643530890345573,
+      "num_tokens": 1754431.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.4913890836543815,
+      "eval_loss": 0.46917298436164856,
+      "eval_mean_token_accuracy": 0.862456685235334,
+      "eval_num_tokens": 1754431.0,
+      "eval_runtime": 96.875,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 720
+    },
+    {
+      "entropy": 0.48629910722374914,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.2625334560871124,
+      "learning_rate": 9.696669560451573e-05,
+      "loss": 0.4475870132446289,
+      "mean_token_accuracy": 0.8631389886140823,
+      "num_tokens": 1804368.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.4911207033451213,
+      "eval_loss": 0.46702805161476135,
+      "eval_mean_token_accuracy": 0.8621654281782549,
+      "eval_num_tokens": 1804368.0,
+      "eval_runtime": 96.8685,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 740
+    },
+    {
+      "entropy": 0.475378530472517,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.2406812161207199,
+      "learning_rate": 9.671139706710683e-05,
+      "loss": 0.4375624179840088,
+      "mean_token_accuracy": 0.8660449728369712,
+      "num_tokens": 1850859.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.48983157044926356,
+      "eval_loss": 0.4654311239719391,
+      "eval_mean_token_accuracy": 0.8626760878535205,
+      "eval_num_tokens": 1850859.0,
+      "eval_runtime": 96.8581,
+      "eval_samples_per_second": 14.206,
+      "eval_steps_per_second": 1.776,
+      "step": 760
+    },
+    {
+      "entropy": 0.4725016601383686,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.2215867042541504,
+      "learning_rate": 9.644187114600152e-05,
+      "loss": 0.44096899032592773,
+      "mean_token_accuracy": 0.8674400866031646,
+      "num_tokens": 1901155.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.4720242848229963,
+      "eval_loss": 0.46539217233657837,
+      "eval_mean_token_accuracy": 0.8630018580791562,
+      "eval_num_tokens": 1901155.0,
+      "eval_runtime": 96.8222,
+      "eval_samples_per_second": 14.212,
+      "eval_steps_per_second": 1.776,
+      "step": 780
+    },
+    {
+      "entropy": 0.4846533551812172,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.2640039920806885,
+      "learning_rate": 9.615819912654142e-05,
+      "loss": 0.451138973236084,
+      "mean_token_accuracy": 0.8665898993611336,
+      "num_tokens": 1949864.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.49363225871740385,
+      "eval_loss": 0.463025838136673,
+      "eval_mean_token_accuracy": 0.8635496410519578,
+      "eval_num_tokens": 1949864.0,
+      "eval_runtime": 96.8241,
+      "eval_samples_per_second": 14.211,
+      "eval_steps_per_second": 1.776,
+      "step": 800
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.9968073072850944e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7bd8b10ce499b826153d768f9491af28a0ff21d5
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/trainer_state.json
@@ -0,0 +1,895 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.0398505603985058,
+  "eval_steps": 20,
+  "global_step": 820,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    },
+    {
+      "entropy": 0.49362134337425234,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.26279115676879883,
+      "learning_rate": 9.819570324736658e-05,
+      "loss": 0.45868754386901855,
+      "mean_token_accuracy": 0.8646741047501564,
+      "num_tokens": 1511476.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.4934931300407232,
+      "eval_loss": 0.47820955514907837,
+      "eval_mean_token_accuracy": 0.8611027102830798,
+      "eval_num_tokens": 1511476.0,
+      "eval_runtime": 96.865,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 620
+    },
+    {
+      "entropy": 0.498298654705286,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.2743106484413147,
+      "learning_rate": 9.802723416796326e-05,
+      "loss": 0.4625910758972168,
+      "mean_token_accuracy": 0.8644399270415306,
+      "num_tokens": 1560069.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.49935419944136644,
+      "eval_loss": 0.4788094460964203,
+      "eval_mean_token_accuracy": 0.8595240989396739,
+      "eval_num_tokens": 1560069.0,
+      "eval_runtime": 96.8858,
+      "eval_samples_per_second": 14.202,
+      "eval_steps_per_second": 1.775,
+      "step": 640
+    },
+    {
+      "entropy": 0.5169856667518615,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.2438548356294632,
+      "learning_rate": 9.784414086634099e-05,
+      "loss": 0.4727236270904541,
+      "mean_token_accuracy": 0.8596569061279297,
+      "num_tokens": 1603766.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.48505116392706715,
+      "eval_loss": 0.4763250946998596,
+      "eval_mean_token_accuracy": 0.8609154699153678,
+      "eval_num_tokens": 1603766.0,
+      "eval_runtime": 96.8748,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 660
+    },
+    {
+      "entropy": 0.48837797343730927,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.2245505303144455,
+      "learning_rate": 9.764647856094226e-05,
+      "loss": 0.45554113388061523,
+      "mean_token_accuracy": 0.8651937782764435,
+      "num_tokens": 1654672.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.47867353970921317,
+      "eval_loss": 0.47178956866264343,
+      "eval_mean_token_accuracy": 0.8621448333873305,
+      "eval_num_tokens": 1654672.0,
+      "eval_runtime": 96.8536,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 680
+    },
+    {
+      "entropy": 0.4763235367834568,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.24530072510242462,
+      "learning_rate": 9.743430686402244e-05,
+      "loss": 0.44700350761413576,
+      "mean_token_accuracy": 0.8670215323567391,
+      "num_tokens": 1706860.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5032368693933931,
+      "eval_loss": 0.466962605714798,
+      "eval_mean_token_accuracy": 0.8626160392927569,
+      "eval_num_tokens": 1706860.0,
+      "eval_runtime": 96.8466,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 700
+    },
+    {
+      "entropy": 0.48205163702368736,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.335552453994751,
+      "learning_rate": 9.720768976367141e-05,
+      "loss": 0.4490774631500244,
+      "mean_token_accuracy": 0.8643530890345573,
+      "num_tokens": 1754431.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.4913890836543815,
+      "eval_loss": 0.46917298436164856,
+      "eval_mean_token_accuracy": 0.862456685235334,
+      "eval_num_tokens": 1754431.0,
+      "eval_runtime": 96.875,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 720
+    },
+    {
+      "entropy": 0.48629910722374914,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.2625334560871124,
+      "learning_rate": 9.696669560451573e-05,
+      "loss": 0.4475870132446289,
+      "mean_token_accuracy": 0.8631389886140823,
+      "num_tokens": 1804368.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.4911207033451213,
+      "eval_loss": 0.46702805161476135,
+      "eval_mean_token_accuracy": 0.8621654281782549,
+      "eval_num_tokens": 1804368.0,
+      "eval_runtime": 96.8685,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 740
+    },
+    {
+      "entropy": 0.475378530472517,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.2406812161207199,
+      "learning_rate": 9.671139706710683e-05,
+      "loss": 0.4375624179840088,
+      "mean_token_accuracy": 0.8660449728369712,
+      "num_tokens": 1850859.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.48983157044926356,
+      "eval_loss": 0.4654311239719391,
+      "eval_mean_token_accuracy": 0.8626760878535205,
+      "eval_num_tokens": 1850859.0,
+      "eval_runtime": 96.8581,
+      "eval_samples_per_second": 14.206,
+      "eval_steps_per_second": 1.776,
+      "step": 760
+    },
+    {
+      "entropy": 0.4725016601383686,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.2215867042541504,
+      "learning_rate": 9.644187114600152e-05,
+      "loss": 0.44096899032592773,
+      "mean_token_accuracy": 0.8674400866031646,
+      "num_tokens": 1901155.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.4720242848229963,
+      "eval_loss": 0.46539217233657837,
+      "eval_mean_token_accuracy": 0.8630018580791562,
+      "eval_num_tokens": 1901155.0,
+      "eval_runtime": 96.8222,
+      "eval_samples_per_second": 14.212,
+      "eval_steps_per_second": 1.776,
+      "step": 780
+    },
+    {
+      "entropy": 0.4846533551812172,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.2640039920806885,
+      "learning_rate": 9.615819912654142e-05,
+      "loss": 0.451138973236084,
+      "mean_token_accuracy": 0.8665898993611336,
+      "num_tokens": 1949864.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.49363225871740385,
+      "eval_loss": 0.463025838136673,
+      "eval_mean_token_accuracy": 0.8635496410519578,
+      "eval_num_tokens": 1949864.0,
+      "eval_runtime": 96.8241,
+      "eval_samples_per_second": 14.211,
+      "eval_steps_per_second": 1.776,
+      "step": 800
+    },
+    {
+      "entropy": 0.4358747288202628,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.28732356429100037,
+      "learning_rate": 9.586046656033853e-05,
+      "loss": 0.39547598361968994,
+      "mean_token_accuracy": 0.8815080141409849,
+      "num_tokens": 1998717.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.4406000122774479,
+      "eval_loss": 0.4686899781227112,
+      "eval_mean_token_accuracy": 0.8642892743958983,
+      "eval_num_tokens": 1998717.0,
+      "eval_runtime": 96.8701,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 820
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.0910259916087296e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..0cc79de8aae56a725cba3c1039fb989f2eaa1b7d
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/trainer_state.json
@@ -0,0 +1,916 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.0896637608966375,
+  "eval_steps": 20,
+  "global_step": 840,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    },
+    {
+      "entropy": 0.49362134337425234,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.26279115676879883,
+      "learning_rate": 9.819570324736658e-05,
+      "loss": 0.45868754386901855,
+      "mean_token_accuracy": 0.8646741047501564,
+      "num_tokens": 1511476.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.4934931300407232,
+      "eval_loss": 0.47820955514907837,
+      "eval_mean_token_accuracy": 0.8611027102830798,
+      "eval_num_tokens": 1511476.0,
+      "eval_runtime": 96.865,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 620
+    },
+    {
+      "entropy": 0.498298654705286,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.2743106484413147,
+      "learning_rate": 9.802723416796326e-05,
+      "loss": 0.4625910758972168,
+      "mean_token_accuracy": 0.8644399270415306,
+      "num_tokens": 1560069.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.49935419944136644,
+      "eval_loss": 0.4788094460964203,
+      "eval_mean_token_accuracy": 0.8595240989396739,
+      "eval_num_tokens": 1560069.0,
+      "eval_runtime": 96.8858,
+      "eval_samples_per_second": 14.202,
+      "eval_steps_per_second": 1.775,
+      "step": 640
+    },
+    {
+      "entropy": 0.5169856667518615,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.2438548356294632,
+      "learning_rate": 9.784414086634099e-05,
+      "loss": 0.4727236270904541,
+      "mean_token_accuracy": 0.8596569061279297,
+      "num_tokens": 1603766.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.48505116392706715,
+      "eval_loss": 0.4763250946998596,
+      "eval_mean_token_accuracy": 0.8609154699153678,
+      "eval_num_tokens": 1603766.0,
+      "eval_runtime": 96.8748,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 660
+    },
+    {
+      "entropy": 0.48837797343730927,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.2245505303144455,
+      "learning_rate": 9.764647856094226e-05,
+      "loss": 0.45554113388061523,
+      "mean_token_accuracy": 0.8651937782764435,
+      "num_tokens": 1654672.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.47867353970921317,
+      "eval_loss": 0.47178956866264343,
+      "eval_mean_token_accuracy": 0.8621448333873305,
+      "eval_num_tokens": 1654672.0,
+      "eval_runtime": 96.8536,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 680
+    },
+    {
+      "entropy": 0.4763235367834568,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.24530072510242462,
+      "learning_rate": 9.743430686402244e-05,
+      "loss": 0.44700350761413576,
+      "mean_token_accuracy": 0.8670215323567391,
+      "num_tokens": 1706860.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5032368693933931,
+      "eval_loss": 0.466962605714798,
+      "eval_mean_token_accuracy": 0.8626160392927569,
+      "eval_num_tokens": 1706860.0,
+      "eval_runtime": 96.8466,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 700
+    },
+    {
+      "entropy": 0.48205163702368736,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.335552453994751,
+      "learning_rate": 9.720768976367141e-05,
+      "loss": 0.4490774631500244,
+      "mean_token_accuracy": 0.8643530890345573,
+      "num_tokens": 1754431.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.4913890836543815,
+      "eval_loss": 0.46917298436164856,
+      "eval_mean_token_accuracy": 0.862456685235334,
+      "eval_num_tokens": 1754431.0,
+      "eval_runtime": 96.875,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 720
+    },
+    {
+      "entropy": 0.48629910722374914,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.2625334560871124,
+      "learning_rate": 9.696669560451573e-05,
+      "loss": 0.4475870132446289,
+      "mean_token_accuracy": 0.8631389886140823,
+      "num_tokens": 1804368.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.4911207033451213,
+      "eval_loss": 0.46702805161476135,
+      "eval_mean_token_accuracy": 0.8621654281782549,
+      "eval_num_tokens": 1804368.0,
+      "eval_runtime": 96.8685,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 740
+    },
+    {
+      "entropy": 0.475378530472517,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.2406812161207199,
+      "learning_rate": 9.671139706710683e-05,
+      "loss": 0.4375624179840088,
+      "mean_token_accuracy": 0.8660449728369712,
+      "num_tokens": 1850859.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.48983157044926356,
+      "eval_loss": 0.4654311239719391,
+      "eval_mean_token_accuracy": 0.8626760878535205,
+      "eval_num_tokens": 1850859.0,
+      "eval_runtime": 96.8581,
+      "eval_samples_per_second": 14.206,
+      "eval_steps_per_second": 1.776,
+      "step": 760
+    },
+    {
+      "entropy": 0.4725016601383686,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.2215867042541504,
+      "learning_rate": 9.644187114600152e-05,
+      "loss": 0.44096899032592773,
+      "mean_token_accuracy": 0.8674400866031646,
+      "num_tokens": 1901155.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.4720242848229963,
+      "eval_loss": 0.46539217233657837,
+      "eval_mean_token_accuracy": 0.8630018580791562,
+      "eval_num_tokens": 1901155.0,
+      "eval_runtime": 96.8222,
+      "eval_samples_per_second": 14.212,
+      "eval_steps_per_second": 1.776,
+      "step": 780
+    },
+    {
+      "entropy": 0.4846533551812172,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.2640039920806885,
+      "learning_rate": 9.615819912654142e-05,
+      "loss": 0.451138973236084,
+      "mean_token_accuracy": 0.8665898993611336,
+      "num_tokens": 1949864.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.49363225871740385,
+      "eval_loss": 0.463025838136673,
+      "eval_mean_token_accuracy": 0.8635496410519578,
+      "eval_num_tokens": 1949864.0,
+      "eval_runtime": 96.8241,
+      "eval_samples_per_second": 14.211,
+      "eval_steps_per_second": 1.776,
+      "step": 800
+    },
+    {
+      "entropy": 0.4358747288202628,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.28732356429100037,
+      "learning_rate": 9.586046656033853e-05,
+      "loss": 0.39547598361968994,
+      "mean_token_accuracy": 0.8815080141409849,
+      "num_tokens": 1998717.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.4406000122774479,
+      "eval_loss": 0.4686899781227112,
+      "eval_mean_token_accuracy": 0.8642892743958983,
+      "eval_num_tokens": 1998717.0,
+      "eval_runtime": 96.8701,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 820
+    },
+    {
+      "entropy": 0.42045016661286355,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.2728928029537201,
+      "learning_rate": 9.554876323947397e-05,
+      "loss": 0.3843829154968262,
+      "mean_token_accuracy": 0.879941563308239,
+      "num_tokens": 2050953.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.4346749455083248,
+      "eval_loss": 0.4729752540588379,
+      "eval_mean_token_accuracy": 0.8633614282968433,
+      "eval_num_tokens": 2050953.0,
+      "eval_runtime": 96.8024,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 840
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.1962220324806656e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e7140daa10fa6f3cc93a2777404f27f2b565aadc
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/trainer_state.json
@@ -0,0 +1,937 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.1394769613947697,
+  "eval_steps": 20,
+  "global_step": 860,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    },
+    {
+      "entropy": 0.49362134337425234,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.26279115676879883,
+      "learning_rate": 9.819570324736658e-05,
+      "loss": 0.45868754386901855,
+      "mean_token_accuracy": 0.8646741047501564,
+      "num_tokens": 1511476.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.4934931300407232,
+      "eval_loss": 0.47820955514907837,
+      "eval_mean_token_accuracy": 0.8611027102830798,
+      "eval_num_tokens": 1511476.0,
+      "eval_runtime": 96.865,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 620
+    },
+    {
+      "entropy": 0.498298654705286,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.2743106484413147,
+      "learning_rate": 9.802723416796326e-05,
+      "loss": 0.4625910758972168,
+      "mean_token_accuracy": 0.8644399270415306,
+      "num_tokens": 1560069.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.49935419944136644,
+      "eval_loss": 0.4788094460964203,
+      "eval_mean_token_accuracy": 0.8595240989396739,
+      "eval_num_tokens": 1560069.0,
+      "eval_runtime": 96.8858,
+      "eval_samples_per_second": 14.202,
+      "eval_steps_per_second": 1.775,
+      "step": 640
+    },
+    {
+      "entropy": 0.5169856667518615,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.2438548356294632,
+      "learning_rate": 9.784414086634099e-05,
+      "loss": 0.4727236270904541,
+      "mean_token_accuracy": 0.8596569061279297,
+      "num_tokens": 1603766.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.48505116392706715,
+      "eval_loss": 0.4763250946998596,
+      "eval_mean_token_accuracy": 0.8609154699153678,
+      "eval_num_tokens": 1603766.0,
+      "eval_runtime": 96.8748,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 660
+    },
+    {
+      "entropy": 0.48837797343730927,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.2245505303144455,
+      "learning_rate": 9.764647856094226e-05,
+      "loss": 0.45554113388061523,
+      "mean_token_accuracy": 0.8651937782764435,
+      "num_tokens": 1654672.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.47867353970921317,
+      "eval_loss": 0.47178956866264343,
+      "eval_mean_token_accuracy": 0.8621448333873305,
+      "eval_num_tokens": 1654672.0,
+      "eval_runtime": 96.8536,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 680
+    },
+    {
+      "entropy": 0.4763235367834568,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.24530072510242462,
+      "learning_rate": 9.743430686402244e-05,
+      "loss": 0.44700350761413576,
+      "mean_token_accuracy": 0.8670215323567391,
+      "num_tokens": 1706860.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5032368693933931,
+      "eval_loss": 0.466962605714798,
+      "eval_mean_token_accuracy": 0.8626160392927569,
+      "eval_num_tokens": 1706860.0,
+      "eval_runtime": 96.8466,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 700
+    },
+    {
+      "entropy": 0.48205163702368736,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.335552453994751,
+      "learning_rate": 9.720768976367141e-05,
+      "loss": 0.4490774631500244,
+      "mean_token_accuracy": 0.8643530890345573,
+      "num_tokens": 1754431.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.4913890836543815,
+      "eval_loss": 0.46917298436164856,
+      "eval_mean_token_accuracy": 0.862456685235334,
+      "eval_num_tokens": 1754431.0,
+      "eval_runtime": 96.875,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 720
+    },
+    {
+      "entropy": 0.48629910722374914,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.2625334560871124,
+      "learning_rate": 9.696669560451573e-05,
+      "loss": 0.4475870132446289,
+      "mean_token_accuracy": 0.8631389886140823,
+      "num_tokens": 1804368.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.4911207033451213,
+      "eval_loss": 0.46702805161476135,
+      "eval_mean_token_accuracy": 0.8621654281782549,
+      "eval_num_tokens": 1804368.0,
+      "eval_runtime": 96.8685,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 740
+    },
+    {
+      "entropy": 0.475378530472517,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.2406812161207199,
+      "learning_rate": 9.671139706710683e-05,
+      "loss": 0.4375624179840088,
+      "mean_token_accuracy": 0.8660449728369712,
+      "num_tokens": 1850859.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.48983157044926356,
+      "eval_loss": 0.4654311239719391,
+      "eval_mean_token_accuracy": 0.8626760878535205,
+      "eval_num_tokens": 1850859.0,
+      "eval_runtime": 96.8581,
+      "eval_samples_per_second": 14.206,
+      "eval_steps_per_second": 1.776,
+      "step": 760
+    },
+    {
+      "entropy": 0.4725016601383686,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.2215867042541504,
+      "learning_rate": 9.644187114600152e-05,
+      "loss": 0.44096899032592773,
+      "mean_token_accuracy": 0.8674400866031646,
+      "num_tokens": 1901155.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.4720242848229963,
+      "eval_loss": 0.46539217233657837,
+      "eval_mean_token_accuracy": 0.8630018580791562,
+      "eval_num_tokens": 1901155.0,
+      "eval_runtime": 96.8222,
+      "eval_samples_per_second": 14.212,
+      "eval_steps_per_second": 1.776,
+      "step": 780
+    },
+    {
+      "entropy": 0.4846533551812172,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.2640039920806885,
+      "learning_rate": 9.615819912654142e-05,
+      "loss": 0.451138973236084,
+      "mean_token_accuracy": 0.8665898993611336,
+      "num_tokens": 1949864.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.49363225871740385,
+      "eval_loss": 0.463025838136673,
+      "eval_mean_token_accuracy": 0.8635496410519578,
+      "eval_num_tokens": 1949864.0,
+      "eval_runtime": 96.8241,
+      "eval_samples_per_second": 14.211,
+      "eval_steps_per_second": 1.776,
+      "step": 800
+    },
+    {
+      "entropy": 0.4358747288202628,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.28732356429100037,
+      "learning_rate": 9.586046656033853e-05,
+      "loss": 0.39547598361968994,
+      "mean_token_accuracy": 0.8815080141409849,
+      "num_tokens": 1998717.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.4406000122774479,
+      "eval_loss": 0.4686899781227112,
+      "eval_mean_token_accuracy": 0.8642892743958983,
+      "eval_num_tokens": 1998717.0,
+      "eval_runtime": 96.8701,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 820
+    },
+    {
+      "entropy": 0.42045016661286355,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.2728928029537201,
+      "learning_rate": 9.554876323947397e-05,
+      "loss": 0.3843829154968262,
+      "mean_token_accuracy": 0.879941563308239,
+      "num_tokens": 2050953.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.4346749455083248,
+      "eval_loss": 0.4729752540588379,
+      "eval_mean_token_accuracy": 0.8633614282968433,
+      "eval_num_tokens": 2050953.0,
+      "eval_runtime": 96.8024,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 840
+    },
+    {
+      "entropy": 0.4300770156085491,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.265896737575531,
+      "learning_rate": 9.522318316941786e-05,
+      "loss": 0.39586963653564455,
+      "mean_token_accuracy": 0.8787443146109581,
+      "num_tokens": 2096118.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4557679929705553,
+      "eval_loss": 0.47163790464401245,
+      "eval_mean_token_accuracy": 0.8636134530222693,
+      "eval_num_tokens": 2096118.0,
+      "eval_runtime": 96.8457,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 860
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.2904389763006464e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..11faea3443d5900ca86a3f97b2ae09d7c6a90278
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/trainer_state.json
@@ -0,0 +1,958 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.1892901618929015,
+  "eval_steps": 20,
+  "global_step": 880,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    },
+    {
+      "entropy": 0.49362134337425234,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.26279115676879883,
+      "learning_rate": 9.819570324736658e-05,
+      "loss": 0.45868754386901855,
+      "mean_token_accuracy": 0.8646741047501564,
+      "num_tokens": 1511476.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.4934931300407232,
+      "eval_loss": 0.47820955514907837,
+      "eval_mean_token_accuracy": 0.8611027102830798,
+      "eval_num_tokens": 1511476.0,
+      "eval_runtime": 96.865,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 620
+    },
+    {
+      "entropy": 0.498298654705286,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.2743106484413147,
+      "learning_rate": 9.802723416796326e-05,
+      "loss": 0.4625910758972168,
+      "mean_token_accuracy": 0.8644399270415306,
+      "num_tokens": 1560069.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.49935419944136644,
+      "eval_loss": 0.4788094460964203,
+      "eval_mean_token_accuracy": 0.8595240989396739,
+      "eval_num_tokens": 1560069.0,
+      "eval_runtime": 96.8858,
+      "eval_samples_per_second": 14.202,
+      "eval_steps_per_second": 1.775,
+      "step": 640
+    },
+    {
+      "entropy": 0.5169856667518615,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.2438548356294632,
+      "learning_rate": 9.784414086634099e-05,
+      "loss": 0.4727236270904541,
+      "mean_token_accuracy": 0.8596569061279297,
+      "num_tokens": 1603766.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.48505116392706715,
+      "eval_loss": 0.4763250946998596,
+      "eval_mean_token_accuracy": 0.8609154699153678,
+      "eval_num_tokens": 1603766.0,
+      "eval_runtime": 96.8748,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 660
+    },
+    {
+      "entropy": 0.48837797343730927,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.2245505303144455,
+      "learning_rate": 9.764647856094226e-05,
+      "loss": 0.45554113388061523,
+      "mean_token_accuracy": 0.8651937782764435,
+      "num_tokens": 1654672.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.47867353970921317,
+      "eval_loss": 0.47178956866264343,
+      "eval_mean_token_accuracy": 0.8621448333873305,
+      "eval_num_tokens": 1654672.0,
+      "eval_runtime": 96.8536,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 680
+    },
+    {
+      "entropy": 0.4763235367834568,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.24530072510242462,
+      "learning_rate": 9.743430686402244e-05,
+      "loss": 0.44700350761413576,
+      "mean_token_accuracy": 0.8670215323567391,
+      "num_tokens": 1706860.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5032368693933931,
+      "eval_loss": 0.466962605714798,
+      "eval_mean_token_accuracy": 0.8626160392927569,
+      "eval_num_tokens": 1706860.0,
+      "eval_runtime": 96.8466,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 700
+    },
+    {
+      "entropy": 0.48205163702368736,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.335552453994751,
+      "learning_rate": 9.720768976367141e-05,
+      "loss": 0.4490774631500244,
+      "mean_token_accuracy": 0.8643530890345573,
+      "num_tokens": 1754431.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.4913890836543815,
+      "eval_loss": 0.46917298436164856,
+      "eval_mean_token_accuracy": 0.862456685235334,
+      "eval_num_tokens": 1754431.0,
+      "eval_runtime": 96.875,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 720
+    },
+    {
+      "entropy": 0.48629910722374914,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.2625334560871124,
+      "learning_rate": 9.696669560451573e-05,
+      "loss": 0.4475870132446289,
+      "mean_token_accuracy": 0.8631389886140823,
+      "num_tokens": 1804368.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.4911207033451213,
+      "eval_loss": 0.46702805161476135,
+      "eval_mean_token_accuracy": 0.8621654281782549,
+      "eval_num_tokens": 1804368.0,
+      "eval_runtime": 96.8685,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 740
+    },
+    {
+      "entropy": 0.475378530472517,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.2406812161207199,
+      "learning_rate": 9.671139706710683e-05,
+      "loss": 0.4375624179840088,
+      "mean_token_accuracy": 0.8660449728369712,
+      "num_tokens": 1850859.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.48983157044926356,
+      "eval_loss": 0.4654311239719391,
+      "eval_mean_token_accuracy": 0.8626760878535205,
+      "eval_num_tokens": 1850859.0,
+      "eval_runtime": 96.8581,
+      "eval_samples_per_second": 14.206,
+      "eval_steps_per_second": 1.776,
+      "step": 760
+    },
+    {
+      "entropy": 0.4725016601383686,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.2215867042541504,
+      "learning_rate": 9.644187114600152e-05,
+      "loss": 0.44096899032592773,
+      "mean_token_accuracy": 0.8674400866031646,
+      "num_tokens": 1901155.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.4720242848229963,
+      "eval_loss": 0.46539217233657837,
+      "eval_mean_token_accuracy": 0.8630018580791562,
+      "eval_num_tokens": 1901155.0,
+      "eval_runtime": 96.8222,
+      "eval_samples_per_second": 14.212,
+      "eval_steps_per_second": 1.776,
+      "step": 780
+    },
+    {
+      "entropy": 0.4846533551812172,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.2640039920806885,
+      "learning_rate": 9.615819912654142e-05,
+      "loss": 0.451138973236084,
+      "mean_token_accuracy": 0.8665898993611336,
+      "num_tokens": 1949864.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.49363225871740385,
+      "eval_loss": 0.463025838136673,
+      "eval_mean_token_accuracy": 0.8635496410519578,
+      "eval_num_tokens": 1949864.0,
+      "eval_runtime": 96.8241,
+      "eval_samples_per_second": 14.211,
+      "eval_steps_per_second": 1.776,
+      "step": 800
+    },
+    {
+      "entropy": 0.4358747288202628,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.28732356429100037,
+      "learning_rate": 9.586046656033853e-05,
+      "loss": 0.39547598361968994,
+      "mean_token_accuracy": 0.8815080141409849,
+      "num_tokens": 1998717.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.4406000122774479,
+      "eval_loss": 0.4686899781227112,
+      "eval_mean_token_accuracy": 0.8642892743958983,
+      "eval_num_tokens": 1998717.0,
+      "eval_runtime": 96.8701,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 820
+    },
+    {
+      "entropy": 0.42045016661286355,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.2728928029537201,
+      "learning_rate": 9.554876323947397e-05,
+      "loss": 0.3843829154968262,
+      "mean_token_accuracy": 0.879941563308239,
+      "num_tokens": 2050953.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.4346749455083248,
+      "eval_loss": 0.4729752540588379,
+      "eval_mean_token_accuracy": 0.8633614282968433,
+      "eval_num_tokens": 2050953.0,
+      "eval_runtime": 96.8024,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 840
+    },
+    {
+      "entropy": 0.4300770156085491,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.265896737575531,
+      "learning_rate": 9.522318316941786e-05,
+      "loss": 0.39586963653564455,
+      "mean_token_accuracy": 0.8787443146109581,
+      "num_tokens": 2096118.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4557679929705553,
+      "eval_loss": 0.47163790464401245,
+      "eval_mean_token_accuracy": 0.8636134530222693,
+      "eval_num_tokens": 2096118.0,
+      "eval_runtime": 96.8457,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 860
+    },
+    {
+      "entropy": 0.4276770144701004,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.2702751159667969,
+      "learning_rate": 9.488382454067867e-05,
+      "loss": 0.3841589689254761,
+      "mean_token_accuracy": 0.8800378635525703,
+      "num_tokens": 2146168.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4377675843100215,
+      "eval_loss": 0.47373291850090027,
+      "eval_mean_token_accuracy": 0.8632283134515896,
+      "eval_num_tokens": 2146168.0,
+      "eval_runtime": 96.8483,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 880
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.3944166646145024e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..4806eb373db1bf74736dc5d0145520579d17d2a0
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/trainer_state.json
@@ -0,0 +1,979 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.2391033623910337,
+  "eval_steps": 20,
+  "global_step": 900,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    },
+    {
+      "entropy": 0.49362134337425234,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.26279115676879883,
+      "learning_rate": 9.819570324736658e-05,
+      "loss": 0.45868754386901855,
+      "mean_token_accuracy": 0.8646741047501564,
+      "num_tokens": 1511476.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.4934931300407232,
+      "eval_loss": 0.47820955514907837,
+      "eval_mean_token_accuracy": 0.8611027102830798,
+      "eval_num_tokens": 1511476.0,
+      "eval_runtime": 96.865,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 620
+    },
+    {
+      "entropy": 0.498298654705286,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.2743106484413147,
+      "learning_rate": 9.802723416796326e-05,
+      "loss": 0.4625910758972168,
+      "mean_token_accuracy": 0.8644399270415306,
+      "num_tokens": 1560069.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.49935419944136644,
+      "eval_loss": 0.4788094460964203,
+      "eval_mean_token_accuracy": 0.8595240989396739,
+      "eval_num_tokens": 1560069.0,
+      "eval_runtime": 96.8858,
+      "eval_samples_per_second": 14.202,
+      "eval_steps_per_second": 1.775,
+      "step": 640
+    },
+    {
+      "entropy": 0.5169856667518615,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.2438548356294632,
+      "learning_rate": 9.784414086634099e-05,
+      "loss": 0.4727236270904541,
+      "mean_token_accuracy": 0.8596569061279297,
+      "num_tokens": 1603766.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.48505116392706715,
+      "eval_loss": 0.4763250946998596,
+      "eval_mean_token_accuracy": 0.8609154699153678,
+      "eval_num_tokens": 1603766.0,
+      "eval_runtime": 96.8748,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 660
+    },
+    {
+      "entropy": 0.48837797343730927,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.2245505303144455,
+      "learning_rate": 9.764647856094226e-05,
+      "loss": 0.45554113388061523,
+      "mean_token_accuracy": 0.8651937782764435,
+      "num_tokens": 1654672.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.47867353970921317,
+      "eval_loss": 0.47178956866264343,
+      "eval_mean_token_accuracy": 0.8621448333873305,
+      "eval_num_tokens": 1654672.0,
+      "eval_runtime": 96.8536,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 680
+    },
+    {
+      "entropy": 0.4763235367834568,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.24530072510242462,
+      "learning_rate": 9.743430686402244e-05,
+      "loss": 0.44700350761413576,
+      "mean_token_accuracy": 0.8670215323567391,
+      "num_tokens": 1706860.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5032368693933931,
+      "eval_loss": 0.466962605714798,
+      "eval_mean_token_accuracy": 0.8626160392927569,
+      "eval_num_tokens": 1706860.0,
+      "eval_runtime": 96.8466,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 700
+    },
+    {
+      "entropy": 0.48205163702368736,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.335552453994751,
+      "learning_rate": 9.720768976367141e-05,
+      "loss": 0.4490774631500244,
+      "mean_token_accuracy": 0.8643530890345573,
+      "num_tokens": 1754431.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.4913890836543815,
+      "eval_loss": 0.46917298436164856,
+      "eval_mean_token_accuracy": 0.862456685235334,
+      "eval_num_tokens": 1754431.0,
+      "eval_runtime": 96.875,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 720
+    },
+    {
+      "entropy": 0.48629910722374914,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.2625334560871124,
+      "learning_rate": 9.696669560451573e-05,
+      "loss": 0.4475870132446289,
+      "mean_token_accuracy": 0.8631389886140823,
+      "num_tokens": 1804368.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.4911207033451213,
+      "eval_loss": 0.46702805161476135,
+      "eval_mean_token_accuracy": 0.8621654281782549,
+      "eval_num_tokens": 1804368.0,
+      "eval_runtime": 96.8685,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 740
+    },
+    {
+      "entropy": 0.475378530472517,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.2406812161207199,
+      "learning_rate": 9.671139706710683e-05,
+      "loss": 0.4375624179840088,
+      "mean_token_accuracy": 0.8660449728369712,
+      "num_tokens": 1850859.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.48983157044926356,
+      "eval_loss": 0.4654311239719391,
+      "eval_mean_token_accuracy": 0.8626760878535205,
+      "eval_num_tokens": 1850859.0,
+      "eval_runtime": 96.8581,
+      "eval_samples_per_second": 14.206,
+      "eval_steps_per_second": 1.776,
+      "step": 760
+    },
+    {
+      "entropy": 0.4725016601383686,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.2215867042541504,
+      "learning_rate": 9.644187114600152e-05,
+      "loss": 0.44096899032592773,
+      "mean_token_accuracy": 0.8674400866031646,
+      "num_tokens": 1901155.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.4720242848229963,
+      "eval_loss": 0.46539217233657837,
+      "eval_mean_token_accuracy": 0.8630018580791562,
+      "eval_num_tokens": 1901155.0,
+      "eval_runtime": 96.8222,
+      "eval_samples_per_second": 14.212,
+      "eval_steps_per_second": 1.776,
+      "step": 780
+    },
+    {
+      "entropy": 0.4846533551812172,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.2640039920806885,
+      "learning_rate": 9.615819912654142e-05,
+      "loss": 0.451138973236084,
+      "mean_token_accuracy": 0.8665898993611336,
+      "num_tokens": 1949864.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.49363225871740385,
+      "eval_loss": 0.463025838136673,
+      "eval_mean_token_accuracy": 0.8635496410519578,
+      "eval_num_tokens": 1949864.0,
+      "eval_runtime": 96.8241,
+      "eval_samples_per_second": 14.211,
+      "eval_steps_per_second": 1.776,
+      "step": 800
+    },
+    {
+      "entropy": 0.4358747288202628,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.28732356429100037,
+      "learning_rate": 9.586046656033853e-05,
+      "loss": 0.39547598361968994,
+      "mean_token_accuracy": 0.8815080141409849,
+      "num_tokens": 1998717.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.4406000122774479,
+      "eval_loss": 0.4686899781227112,
+      "eval_mean_token_accuracy": 0.8642892743958983,
+      "eval_num_tokens": 1998717.0,
+      "eval_runtime": 96.8701,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 820
+    },
+    {
+      "entropy": 0.42045016661286355,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.2728928029537201,
+      "learning_rate": 9.554876323947397e-05,
+      "loss": 0.3843829154968262,
+      "mean_token_accuracy": 0.879941563308239,
+      "num_tokens": 2050953.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.4346749455083248,
+      "eval_loss": 0.4729752540588379,
+      "eval_mean_token_accuracy": 0.8633614282968433,
+      "eval_num_tokens": 2050953.0,
+      "eval_runtime": 96.8024,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 840
+    },
+    {
+      "entropy": 0.4300770156085491,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.265896737575531,
+      "learning_rate": 9.522318316941786e-05,
+      "loss": 0.39586963653564455,
+      "mean_token_accuracy": 0.8787443146109581,
+      "num_tokens": 2096118.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4557679929705553,
+      "eval_loss": 0.47163790464401245,
+      "eval_mean_token_accuracy": 0.8636134530222693,
+      "eval_num_tokens": 2096118.0,
+      "eval_runtime": 96.8457,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 860
+    },
+    {
+      "entropy": 0.4276770144701004,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.2702751159667969,
+      "learning_rate": 9.488382454067867e-05,
+      "loss": 0.3841589689254761,
+      "mean_token_accuracy": 0.8800378635525703,
+      "num_tokens": 2146168.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4377675843100215,
+      "eval_loss": 0.47373291850090027,
+      "eval_mean_token_accuracy": 0.8632283134515896,
+      "eval_num_tokens": 2146168.0,
+      "eval_runtime": 96.8483,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 880
+    },
+    {
+      "entropy": 0.4314340800046921,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.27613386511802673,
+      "learning_rate": 9.453078969919021e-05,
+      "loss": 0.3940398931503296,
+      "mean_token_accuracy": 0.8791197642683983,
+      "num_tokens": 2195193.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4267724165043166,
+      "eval_loss": 0.4770275056362152,
+      "eval_mean_token_accuracy": 0.8623441904783249,
+      "eval_num_tokens": 2195193.0,
+      "eval_runtime": 96.846,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 900
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.490360188059648e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ff1d3172999def5b75f3c059541d50197cbc00d1
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/trainer_state.json
@@ -0,0 +1,1000 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.2889165628891655,
+  "eval_steps": 20,
+  "global_step": 920,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    },
+    {
+      "entropy": 0.49362134337425234,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.26279115676879883,
+      "learning_rate": 9.819570324736658e-05,
+      "loss": 0.45868754386901855,
+      "mean_token_accuracy": 0.8646741047501564,
+      "num_tokens": 1511476.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.4934931300407232,
+      "eval_loss": 0.47820955514907837,
+      "eval_mean_token_accuracy": 0.8611027102830798,
+      "eval_num_tokens": 1511476.0,
+      "eval_runtime": 96.865,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 620
+    },
+    {
+      "entropy": 0.498298654705286,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.2743106484413147,
+      "learning_rate": 9.802723416796326e-05,
+      "loss": 0.4625910758972168,
+      "mean_token_accuracy": 0.8644399270415306,
+      "num_tokens": 1560069.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.49935419944136644,
+      "eval_loss": 0.4788094460964203,
+      "eval_mean_token_accuracy": 0.8595240989396739,
+      "eval_num_tokens": 1560069.0,
+      "eval_runtime": 96.8858,
+      "eval_samples_per_second": 14.202,
+      "eval_steps_per_second": 1.775,
+      "step": 640
+    },
+    {
+      "entropy": 0.5169856667518615,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.2438548356294632,
+      "learning_rate": 9.784414086634099e-05,
+      "loss": 0.4727236270904541,
+      "mean_token_accuracy": 0.8596569061279297,
+      "num_tokens": 1603766.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.48505116392706715,
+      "eval_loss": 0.4763250946998596,
+      "eval_mean_token_accuracy": 0.8609154699153678,
+      "eval_num_tokens": 1603766.0,
+      "eval_runtime": 96.8748,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 660
+    },
+    {
+      "entropy": 0.48837797343730927,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.2245505303144455,
+      "learning_rate": 9.764647856094226e-05,
+      "loss": 0.45554113388061523,
+      "mean_token_accuracy": 0.8651937782764435,
+      "num_tokens": 1654672.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.47867353970921317,
+      "eval_loss": 0.47178956866264343,
+      "eval_mean_token_accuracy": 0.8621448333873305,
+      "eval_num_tokens": 1654672.0,
+      "eval_runtime": 96.8536,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 680
+    },
+    {
+      "entropy": 0.4763235367834568,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.24530072510242462,
+      "learning_rate": 9.743430686402244e-05,
+      "loss": 0.44700350761413576,
+      "mean_token_accuracy": 0.8670215323567391,
+      "num_tokens": 1706860.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5032368693933931,
+      "eval_loss": 0.466962605714798,
+      "eval_mean_token_accuracy": 0.8626160392927569,
+      "eval_num_tokens": 1706860.0,
+      "eval_runtime": 96.8466,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 700
+    },
+    {
+      "entropy": 0.48205163702368736,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.335552453994751,
+      "learning_rate": 9.720768976367141e-05,
+      "loss": 0.4490774631500244,
+      "mean_token_accuracy": 0.8643530890345573,
+      "num_tokens": 1754431.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.4913890836543815,
+      "eval_loss": 0.46917298436164856,
+      "eval_mean_token_accuracy": 0.862456685235334,
+      "eval_num_tokens": 1754431.0,
+      "eval_runtime": 96.875,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 720
+    },
+    {
+      "entropy": 0.48629910722374914,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.2625334560871124,
+      "learning_rate": 9.696669560451573e-05,
+      "loss": 0.4475870132446289,
+      "mean_token_accuracy": 0.8631389886140823,
+      "num_tokens": 1804368.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.4911207033451213,
+      "eval_loss": 0.46702805161476135,
+      "eval_mean_token_accuracy": 0.8621654281782549,
+      "eval_num_tokens": 1804368.0,
+      "eval_runtime": 96.8685,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 740
+    },
+    {
+      "entropy": 0.475378530472517,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.2406812161207199,
+      "learning_rate": 9.671139706710683e-05,
+      "loss": 0.4375624179840088,
+      "mean_token_accuracy": 0.8660449728369712,
+      "num_tokens": 1850859.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.48983157044926356,
+      "eval_loss": 0.4654311239719391,
+      "eval_mean_token_accuracy": 0.8626760878535205,
+      "eval_num_tokens": 1850859.0,
+      "eval_runtime": 96.8581,
+      "eval_samples_per_second": 14.206,
+      "eval_steps_per_second": 1.776,
+      "step": 760
+    },
+    {
+      "entropy": 0.4725016601383686,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.2215867042541504,
+      "learning_rate": 9.644187114600152e-05,
+      "loss": 0.44096899032592773,
+      "mean_token_accuracy": 0.8674400866031646,
+      "num_tokens": 1901155.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.4720242848229963,
+      "eval_loss": 0.46539217233657837,
+      "eval_mean_token_accuracy": 0.8630018580791562,
+      "eval_num_tokens": 1901155.0,
+      "eval_runtime": 96.8222,
+      "eval_samples_per_second": 14.212,
+      "eval_steps_per_second": 1.776,
+      "step": 780
+    },
+    {
+      "entropy": 0.4846533551812172,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.2640039920806885,
+      "learning_rate": 9.615819912654142e-05,
+      "loss": 0.451138973236084,
+      "mean_token_accuracy": 0.8665898993611336,
+      "num_tokens": 1949864.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.49363225871740385,
+      "eval_loss": 0.463025838136673,
+      "eval_mean_token_accuracy": 0.8635496410519578,
+      "eval_num_tokens": 1949864.0,
+      "eval_runtime": 96.8241,
+      "eval_samples_per_second": 14.211,
+      "eval_steps_per_second": 1.776,
+      "step": 800
+    },
+    {
+      "entropy": 0.4358747288202628,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.28732356429100037,
+      "learning_rate": 9.586046656033853e-05,
+      "loss": 0.39547598361968994,
+      "mean_token_accuracy": 0.8815080141409849,
+      "num_tokens": 1998717.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.4406000122774479,
+      "eval_loss": 0.4686899781227112,
+      "eval_mean_token_accuracy": 0.8642892743958983,
+      "eval_num_tokens": 1998717.0,
+      "eval_runtime": 96.8701,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 820
+    },
+    {
+      "entropy": 0.42045016661286355,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.2728928029537201,
+      "learning_rate": 9.554876323947397e-05,
+      "loss": 0.3843829154968262,
+      "mean_token_accuracy": 0.879941563308239,
+      "num_tokens": 2050953.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.4346749455083248,
+      "eval_loss": 0.4729752540588379,
+      "eval_mean_token_accuracy": 0.8633614282968433,
+      "eval_num_tokens": 2050953.0,
+      "eval_runtime": 96.8024,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 840
+    },
+    {
+      "entropy": 0.4300770156085491,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.265896737575531,
+      "learning_rate": 9.522318316941786e-05,
+      "loss": 0.39586963653564455,
+      "mean_token_accuracy": 0.8787443146109581,
+      "num_tokens": 2096118.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4557679929705553,
+      "eval_loss": 0.47163790464401245,
+      "eval_mean_token_accuracy": 0.8636134530222693,
+      "eval_num_tokens": 2096118.0,
+      "eval_runtime": 96.8457,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 860
+    },
+    {
+      "entropy": 0.4276770144701004,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.2702751159667969,
+      "learning_rate": 9.488382454067867e-05,
+      "loss": 0.3841589689254761,
+      "mean_token_accuracy": 0.8800378635525703,
+      "num_tokens": 2146168.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4377675843100215,
+      "eval_loss": 0.47373291850090027,
+      "eval_mean_token_accuracy": 0.8632283134515896,
+      "eval_num_tokens": 2146168.0,
+      "eval_runtime": 96.8483,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 880
+    },
+    {
+      "entropy": 0.4314340800046921,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.27613386511802673,
+      "learning_rate": 9.453078969919021e-05,
+      "loss": 0.3940398931503296,
+      "mean_token_accuracy": 0.8791197642683983,
+      "num_tokens": 2195193.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4267724165043166,
+      "eval_loss": 0.4770275056362152,
+      "eval_mean_token_accuracy": 0.8623441904783249,
+      "eval_num_tokens": 2195193.0,
+      "eval_runtime": 96.846,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 900
+    },
+    {
+      "entropy": 0.41646718233823776,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.34193891286849976,
+      "learning_rate": 9.416418511544553e-05,
+      "loss": 0.3823988676071167,
+      "mean_token_accuracy": 0.881490395963192,
+      "num_tokens": 2241455.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.45252677469059477,
+      "eval_loss": 0.4692145884037018,
+      "eval_mean_token_accuracy": 0.8639034258071766,
+      "eval_num_tokens": 2241455.0,
+      "eval_runtime": 96.7456,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 920
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.5899587691790336e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..4a0b8a344508837d11f2de1bf96d74b9ed74a38a
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/trainer_state.json
@@ -0,0 +1,1021 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.3387297633872977,
+  "eval_steps": 20,
+  "global_step": 940,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    },
+    {
+      "entropy": 0.49362134337425234,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.26279115676879883,
+      "learning_rate": 9.819570324736658e-05,
+      "loss": 0.45868754386901855,
+      "mean_token_accuracy": 0.8646741047501564,
+      "num_tokens": 1511476.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.4934931300407232,
+      "eval_loss": 0.47820955514907837,
+      "eval_mean_token_accuracy": 0.8611027102830798,
+      "eval_num_tokens": 1511476.0,
+      "eval_runtime": 96.865,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 620
+    },
+    {
+      "entropy": 0.498298654705286,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.2743106484413147,
+      "learning_rate": 9.802723416796326e-05,
+      "loss": 0.4625910758972168,
+      "mean_token_accuracy": 0.8644399270415306,
+      "num_tokens": 1560069.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.49935419944136644,
+      "eval_loss": 0.4788094460964203,
+      "eval_mean_token_accuracy": 0.8595240989396739,
+      "eval_num_tokens": 1560069.0,
+      "eval_runtime": 96.8858,
+      "eval_samples_per_second": 14.202,
+      "eval_steps_per_second": 1.775,
+      "step": 640
+    },
+    {
+      "entropy": 0.5169856667518615,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.2438548356294632,
+      "learning_rate": 9.784414086634099e-05,
+      "loss": 0.4727236270904541,
+      "mean_token_accuracy": 0.8596569061279297,
+      "num_tokens": 1603766.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.48505116392706715,
+      "eval_loss": 0.4763250946998596,
+      "eval_mean_token_accuracy": 0.8609154699153678,
+      "eval_num_tokens": 1603766.0,
+      "eval_runtime": 96.8748,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 660
+    },
+    {
+      "entropy": 0.48837797343730927,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.2245505303144455,
+      "learning_rate": 9.764647856094226e-05,
+      "loss": 0.45554113388061523,
+      "mean_token_accuracy": 0.8651937782764435,
+      "num_tokens": 1654672.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.47867353970921317,
+      "eval_loss": 0.47178956866264343,
+      "eval_mean_token_accuracy": 0.8621448333873305,
+      "eval_num_tokens": 1654672.0,
+      "eval_runtime": 96.8536,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 680
+    },
+    {
+      "entropy": 0.4763235367834568,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.24530072510242462,
+      "learning_rate": 9.743430686402244e-05,
+      "loss": 0.44700350761413576,
+      "mean_token_accuracy": 0.8670215323567391,
+      "num_tokens": 1706860.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5032368693933931,
+      "eval_loss": 0.466962605714798,
+      "eval_mean_token_accuracy": 0.8626160392927569,
+      "eval_num_tokens": 1706860.0,
+      "eval_runtime": 96.8466,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 700
+    },
+    {
+      "entropy": 0.48205163702368736,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.335552453994751,
+      "learning_rate": 9.720768976367141e-05,
+      "loss": 0.4490774631500244,
+      "mean_token_accuracy": 0.8643530890345573,
+      "num_tokens": 1754431.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.4913890836543815,
+      "eval_loss": 0.46917298436164856,
+      "eval_mean_token_accuracy": 0.862456685235334,
+      "eval_num_tokens": 1754431.0,
+      "eval_runtime": 96.875,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 720
+    },
+    {
+      "entropy": 0.48629910722374914,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.2625334560871124,
+      "learning_rate": 9.696669560451573e-05,
+      "loss": 0.4475870132446289,
+      "mean_token_accuracy": 0.8631389886140823,
+      "num_tokens": 1804368.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.4911207033451213,
+      "eval_loss": 0.46702805161476135,
+      "eval_mean_token_accuracy": 0.8621654281782549,
+      "eval_num_tokens": 1804368.0,
+      "eval_runtime": 96.8685,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 740
+    },
+    {
+      "entropy": 0.475378530472517,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.2406812161207199,
+      "learning_rate": 9.671139706710683e-05,
+      "loss": 0.4375624179840088,
+      "mean_token_accuracy": 0.8660449728369712,
+      "num_tokens": 1850859.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.48983157044926356,
+      "eval_loss": 0.4654311239719391,
+      "eval_mean_token_accuracy": 0.8626760878535205,
+      "eval_num_tokens": 1850859.0,
+      "eval_runtime": 96.8581,
+      "eval_samples_per_second": 14.206,
+      "eval_steps_per_second": 1.776,
+      "step": 760
+    },
+    {
+      "entropy": 0.4725016601383686,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.2215867042541504,
+      "learning_rate": 9.644187114600152e-05,
+      "loss": 0.44096899032592773,
+      "mean_token_accuracy": 0.8674400866031646,
+      "num_tokens": 1901155.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.4720242848229963,
+      "eval_loss": 0.46539217233657837,
+      "eval_mean_token_accuracy": 0.8630018580791562,
+      "eval_num_tokens": 1901155.0,
+      "eval_runtime": 96.8222,
+      "eval_samples_per_second": 14.212,
+      "eval_steps_per_second": 1.776,
+      "step": 780
+    },
+    {
+      "entropy": 0.4846533551812172,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.2640039920806885,
+      "learning_rate": 9.615819912654142e-05,
+      "loss": 0.451138973236084,
+      "mean_token_accuracy": 0.8665898993611336,
+      "num_tokens": 1949864.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.49363225871740385,
+      "eval_loss": 0.463025838136673,
+      "eval_mean_token_accuracy": 0.8635496410519578,
+      "eval_num_tokens": 1949864.0,
+      "eval_runtime": 96.8241,
+      "eval_samples_per_second": 14.211,
+      "eval_steps_per_second": 1.776,
+      "step": 800
+    },
+    {
+      "entropy": 0.4358747288202628,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.28732356429100037,
+      "learning_rate": 9.586046656033853e-05,
+      "loss": 0.39547598361968994,
+      "mean_token_accuracy": 0.8815080141409849,
+      "num_tokens": 1998717.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.4406000122774479,
+      "eval_loss": 0.4686899781227112,
+      "eval_mean_token_accuracy": 0.8642892743958983,
+      "eval_num_tokens": 1998717.0,
+      "eval_runtime": 96.8701,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 820
+    },
+    {
+      "entropy": 0.42045016661286355,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.2728928029537201,
+      "learning_rate": 9.554876323947397e-05,
+      "loss": 0.3843829154968262,
+      "mean_token_accuracy": 0.879941563308239,
+      "num_tokens": 2050953.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.4346749455083248,
+      "eval_loss": 0.4729752540588379,
+      "eval_mean_token_accuracy": 0.8633614282968433,
+      "eval_num_tokens": 2050953.0,
+      "eval_runtime": 96.8024,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 840
+    },
+    {
+      "entropy": 0.4300770156085491,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.265896737575531,
+      "learning_rate": 9.522318316941786e-05,
+      "loss": 0.39586963653564455,
+      "mean_token_accuracy": 0.8787443146109581,
+      "num_tokens": 2096118.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4557679929705553,
+      "eval_loss": 0.47163790464401245,
+      "eval_mean_token_accuracy": 0.8636134530222693,
+      "eval_num_tokens": 2096118.0,
+      "eval_runtime": 96.8457,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 860
+    },
+    {
+      "entropy": 0.4276770144701004,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.2702751159667969,
+      "learning_rate": 9.488382454067867e-05,
+      "loss": 0.3841589689254761,
+      "mean_token_accuracy": 0.8800378635525703,
+      "num_tokens": 2146168.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4377675843100215,
+      "eval_loss": 0.47373291850090027,
+      "eval_mean_token_accuracy": 0.8632283134515896,
+      "eval_num_tokens": 2146168.0,
+      "eval_runtime": 96.8483,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 880
+    },
+    {
+      "entropy": 0.4314340800046921,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.27613386511802673,
+      "learning_rate": 9.453078969919021e-05,
+      "loss": 0.3940398931503296,
+      "mean_token_accuracy": 0.8791197642683983,
+      "num_tokens": 2195193.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4267724165043166,
+      "eval_loss": 0.4770275056362152,
+      "eval_mean_token_accuracy": 0.8623441904783249,
+      "eval_num_tokens": 2195193.0,
+      "eval_runtime": 96.846,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 900
+    },
+    {
+      "entropy": 0.41646718233823776,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.34193891286849976,
+      "learning_rate": 9.416418511544553e-05,
+      "loss": 0.3823988676071167,
+      "mean_token_accuracy": 0.881490395963192,
+      "num_tokens": 2241455.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.45252677469059477,
+      "eval_loss": 0.4692145884037018,
+      "eval_mean_token_accuracy": 0.8639034258071766,
+      "eval_num_tokens": 2241455.0,
+      "eval_runtime": 96.7456,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 920
+    },
+    {
+      "entropy": 0.43459259644150733,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.2719372808933258,
+      "learning_rate": 9.378412135238683e-05,
+      "loss": 0.39447042942047117,
+      "mean_token_accuracy": 0.8789274036884308,
+      "num_tokens": 2286048.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.4377528974829718,
+      "eval_loss": 0.47256559133529663,
+      "eval_mean_token_accuracy": 0.8633056709239649,
+      "eval_num_tokens": 2286048.0,
+      "eval_runtime": 96.9805,
+      "eval_samples_per_second": 14.188,
+      "eval_steps_per_second": 1.774,
+      "step": 940
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.675410536595456e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..8dbc0ef296e41ea063df3f5001c27d08305facd9
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/trainer_state.json
@@ -0,0 +1,1042 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.3885429638854294,
+  "eval_steps": 20,
+  "global_step": 960,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    },
+    {
+      "entropy": 0.49362134337425234,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.26279115676879883,
+      "learning_rate": 9.819570324736658e-05,
+      "loss": 0.45868754386901855,
+      "mean_token_accuracy": 0.8646741047501564,
+      "num_tokens": 1511476.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.4934931300407232,
+      "eval_loss": 0.47820955514907837,
+      "eval_mean_token_accuracy": 0.8611027102830798,
+      "eval_num_tokens": 1511476.0,
+      "eval_runtime": 96.865,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 620
+    },
+    {
+      "entropy": 0.498298654705286,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.2743106484413147,
+      "learning_rate": 9.802723416796326e-05,
+      "loss": 0.4625910758972168,
+      "mean_token_accuracy": 0.8644399270415306,
+      "num_tokens": 1560069.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.49935419944136644,
+      "eval_loss": 0.4788094460964203,
+      "eval_mean_token_accuracy": 0.8595240989396739,
+      "eval_num_tokens": 1560069.0,
+      "eval_runtime": 96.8858,
+      "eval_samples_per_second": 14.202,
+      "eval_steps_per_second": 1.775,
+      "step": 640
+    },
+    {
+      "entropy": 0.5169856667518615,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.2438548356294632,
+      "learning_rate": 9.784414086634099e-05,
+      "loss": 0.4727236270904541,
+      "mean_token_accuracy": 0.8596569061279297,
+      "num_tokens": 1603766.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.48505116392706715,
+      "eval_loss": 0.4763250946998596,
+      "eval_mean_token_accuracy": 0.8609154699153678,
+      "eval_num_tokens": 1603766.0,
+      "eval_runtime": 96.8748,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 660
+    },
+    {
+      "entropy": 0.48837797343730927,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.2245505303144455,
+      "learning_rate": 9.764647856094226e-05,
+      "loss": 0.45554113388061523,
+      "mean_token_accuracy": 0.8651937782764435,
+      "num_tokens": 1654672.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.47867353970921317,
+      "eval_loss": 0.47178956866264343,
+      "eval_mean_token_accuracy": 0.8621448333873305,
+      "eval_num_tokens": 1654672.0,
+      "eval_runtime": 96.8536,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 680
+    },
+    {
+      "entropy": 0.4763235367834568,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.24530072510242462,
+      "learning_rate": 9.743430686402244e-05,
+      "loss": 0.44700350761413576,
+      "mean_token_accuracy": 0.8670215323567391,
+      "num_tokens": 1706860.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5032368693933931,
+      "eval_loss": 0.466962605714798,
+      "eval_mean_token_accuracy": 0.8626160392927569,
+      "eval_num_tokens": 1706860.0,
+      "eval_runtime": 96.8466,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 700
+    },
+    {
+      "entropy": 0.48205163702368736,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.335552453994751,
+      "learning_rate": 9.720768976367141e-05,
+      "loss": 0.4490774631500244,
+      "mean_token_accuracy": 0.8643530890345573,
+      "num_tokens": 1754431.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.4913890836543815,
+      "eval_loss": 0.46917298436164856,
+      "eval_mean_token_accuracy": 0.862456685235334,
+      "eval_num_tokens": 1754431.0,
+      "eval_runtime": 96.875,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 720
+    },
+    {
+      "entropy": 0.48629910722374914,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.2625334560871124,
+      "learning_rate": 9.696669560451573e-05,
+      "loss": 0.4475870132446289,
+      "mean_token_accuracy": 0.8631389886140823,
+      "num_tokens": 1804368.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.4911207033451213,
+      "eval_loss": 0.46702805161476135,
+      "eval_mean_token_accuracy": 0.8621654281782549,
+      "eval_num_tokens": 1804368.0,
+      "eval_runtime": 96.8685,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 740
+    },
+    {
+      "entropy": 0.475378530472517,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.2406812161207199,
+      "learning_rate": 9.671139706710683e-05,
+      "loss": 0.4375624179840088,
+      "mean_token_accuracy": 0.8660449728369712,
+      "num_tokens": 1850859.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.48983157044926356,
+      "eval_loss": 0.4654311239719391,
+      "eval_mean_token_accuracy": 0.8626760878535205,
+      "eval_num_tokens": 1850859.0,
+      "eval_runtime": 96.8581,
+      "eval_samples_per_second": 14.206,
+      "eval_steps_per_second": 1.776,
+      "step": 760
+    },
+    {
+      "entropy": 0.4725016601383686,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.2215867042541504,
+      "learning_rate": 9.644187114600152e-05,
+      "loss": 0.44096899032592773,
+      "mean_token_accuracy": 0.8674400866031646,
+      "num_tokens": 1901155.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.4720242848229963,
+      "eval_loss": 0.46539217233657837,
+      "eval_mean_token_accuracy": 0.8630018580791562,
+      "eval_num_tokens": 1901155.0,
+      "eval_runtime": 96.8222,
+      "eval_samples_per_second": 14.212,
+      "eval_steps_per_second": 1.776,
+      "step": 780
+    },
+    {
+      "entropy": 0.4846533551812172,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.2640039920806885,
+      "learning_rate": 9.615819912654142e-05,
+      "loss": 0.451138973236084,
+      "mean_token_accuracy": 0.8665898993611336,
+      "num_tokens": 1949864.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.49363225871740385,
+      "eval_loss": 0.463025838136673,
+      "eval_mean_token_accuracy": 0.8635496410519578,
+      "eval_num_tokens": 1949864.0,
+      "eval_runtime": 96.8241,
+      "eval_samples_per_second": 14.211,
+      "eval_steps_per_second": 1.776,
+      "step": 800
+    },
+    {
+      "entropy": 0.4358747288202628,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.28732356429100037,
+      "learning_rate": 9.586046656033853e-05,
+      "loss": 0.39547598361968994,
+      "mean_token_accuracy": 0.8815080141409849,
+      "num_tokens": 1998717.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.4406000122774479,
+      "eval_loss": 0.4686899781227112,
+      "eval_mean_token_accuracy": 0.8642892743958983,
+      "eval_num_tokens": 1998717.0,
+      "eval_runtime": 96.8701,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 820
+    },
+    {
+      "entropy": 0.42045016661286355,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.2728928029537201,
+      "learning_rate": 9.554876323947397e-05,
+      "loss": 0.3843829154968262,
+      "mean_token_accuracy": 0.879941563308239,
+      "num_tokens": 2050953.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.4346749455083248,
+      "eval_loss": 0.4729752540588379,
+      "eval_mean_token_accuracy": 0.8633614282968433,
+      "eval_num_tokens": 2050953.0,
+      "eval_runtime": 96.8024,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 840
+    },
+    {
+      "entropy": 0.4300770156085491,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.265896737575531,
+      "learning_rate": 9.522318316941786e-05,
+      "loss": 0.39586963653564455,
+      "mean_token_accuracy": 0.8787443146109581,
+      "num_tokens": 2096118.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4557679929705553,
+      "eval_loss": 0.47163790464401245,
+      "eval_mean_token_accuracy": 0.8636134530222693,
+      "eval_num_tokens": 2096118.0,
+      "eval_runtime": 96.8457,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 860
+    },
+    {
+      "entropy": 0.4276770144701004,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.2702751159667969,
+      "learning_rate": 9.488382454067867e-05,
+      "loss": 0.3841589689254761,
+      "mean_token_accuracy": 0.8800378635525703,
+      "num_tokens": 2146168.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4377675843100215,
+      "eval_loss": 0.47373291850090027,
+      "eval_mean_token_accuracy": 0.8632283134515896,
+      "eval_num_tokens": 2146168.0,
+      "eval_runtime": 96.8483,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 880
+    },
+    {
+      "entropy": 0.4314340800046921,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.27613386511802673,
+      "learning_rate": 9.453078969919021e-05,
+      "loss": 0.3940398931503296,
+      "mean_token_accuracy": 0.8791197642683983,
+      "num_tokens": 2195193.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4267724165043166,
+      "eval_loss": 0.4770275056362152,
+      "eval_mean_token_accuracy": 0.8623441904783249,
+      "eval_num_tokens": 2195193.0,
+      "eval_runtime": 96.846,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 900
+    },
+    {
+      "entropy": 0.41646718233823776,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.34193891286849976,
+      "learning_rate": 9.416418511544553e-05,
+      "loss": 0.3823988676071167,
+      "mean_token_accuracy": 0.881490395963192,
+      "num_tokens": 2241455.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.45252677469059477,
+      "eval_loss": 0.4692145884037018,
+      "eval_mean_token_accuracy": 0.8639034258071766,
+      "eval_num_tokens": 2241455.0,
+      "eval_runtime": 96.7456,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 920
+    },
+    {
+      "entropy": 0.43459259644150733,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.2719372808933258,
+      "learning_rate": 9.378412135238683e-05,
+      "loss": 0.39447042942047117,
+      "mean_token_accuracy": 0.8789274036884308,
+      "num_tokens": 2286048.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.4377528974829718,
+      "eval_loss": 0.47256559133529663,
+      "eval_mean_token_accuracy": 0.8633056709239649,
+      "eval_num_tokens": 2286048.0,
+      "eval_runtime": 96.9805,
+      "eval_samples_per_second": 14.188,
+      "eval_steps_per_second": 1.774,
+      "step": 940
+    },
+    {
+      "entropy": 0.4245278872549534,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.303096741437912,
+      "learning_rate": 9.33907130320613e-05,
+      "loss": 0.3913075923919678,
+      "mean_token_accuracy": 0.8795066639780998,
+      "num_tokens": 2336997.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.43375605102195297,
+      "eval_loss": 0.4710635840892792,
+      "eval_mean_token_accuracy": 0.86402241400508,
+      "eval_num_tokens": 2336997.0,
+      "eval_runtime": 96.8813,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 960
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.777320506567885e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/README.md b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b69fa97fbe239b1f2988d3e366763603689ad4b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3-14B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-14B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986de34fc8ff08011823a65c6de1a8e1fb1fd8c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-14B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.002044103085291549,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c960ecf0d33fd7b8c99d12680c0e74a82b36d446
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/tokenizer_config.json
@@ -0,0 +1,29 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c828657ab3b3b54dc90bb6586ba862e1be1e234c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3-14B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/trainer_state.json
@@ -0,0 +1,1063 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.4383561643835616,
+  "eval_steps": 20,
+  "global_step": 980,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.755050191283226,
+      "epoch": 0.049813200498132,
+      "grad_norm": 0.7545326948165894,
+      "learning_rate": 4.682530224133709e-06,
+      "loss": 1.7114290237426757,
+      "mean_token_accuracy": 0.6533655762672425,
+      "num_tokens": 48808.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.7040477831696355,
+      "eval_loss": 1.679547905921936,
+      "eval_mean_token_accuracy": 0.6626217687545821,
+      "eval_num_tokens": 48808.0,
+      "eval_runtime": 96.7039,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 20
+    },
+    {
+      "entropy": 1.5912526160478593,
+      "epoch": 0.099626400996264,
+      "grad_norm": 0.7139635682106018,
+      "learning_rate": 9.61150940743235e-06,
+      "loss": 1.4942971229553224,
+      "mean_token_accuracy": 0.6889698803424835,
+      "num_tokens": 94643.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 1.2383613600287326,
+      "eval_loss": 1.1722209453582764,
+      "eval_mean_token_accuracy": 0.7378552531780198,
+      "eval_num_tokens": 94643.0,
+      "eval_runtime": 96.7779,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.777,
+      "step": 40
+    },
+    {
+      "entropy": 1.0441708981990814,
+      "epoch": 0.149439601494396,
+      "grad_norm": 0.5701689124107361,
+      "learning_rate": 1.454048859073099e-05,
+      "loss": 0.9424296379089355,
+      "mean_token_accuracy": 0.7681305900216102,
+      "num_tokens": 143425.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.8593970166389332,
+      "eval_loss": 0.7761058211326599,
+      "eval_mean_token_accuracy": 0.7953345820654271,
+      "eval_num_tokens": 143425.0,
+      "eval_runtime": 96.7416,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 60
+    },
+    {
+      "entropy": 0.7493324965238571,
+      "epoch": 0.199252801992528,
+      "grad_norm": 0.5119306445121765,
+      "learning_rate": 1.946946777402963e-05,
+      "loss": 0.703174638748169,
+      "mean_token_accuracy": 0.8108513653278351,
+      "num_tokens": 195524.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6977347970701927,
+      "eval_loss": 0.6606373190879822,
+      "eval_mean_token_accuracy": 0.8206594520530035,
+      "eval_num_tokens": 195524.0,
+      "eval_runtime": 96.7013,
+      "eval_samples_per_second": 14.229,
+      "eval_steps_per_second": 1.779,
+      "step": 80
+    },
+    {
+      "entropy": 0.6704841658473015,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.397317498922348,
+      "learning_rate": 2.439844695732827e-05,
+      "loss": 0.6182228088378906,
+      "mean_token_accuracy": 0.8291108086705208,
+      "num_tokens": 244338.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6474089567051378,
+      "eval_loss": 0.6101279854774475,
+      "eval_mean_token_accuracy": 0.8300852463688961,
+      "eval_num_tokens": 244338.0,
+      "eval_runtime": 96.6655,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 100
+    },
+    {
+      "entropy": 0.6319684013724327,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.48044654726982117,
+      "learning_rate": 2.932742614062691e-05,
+      "loss": 0.5982244491577149,
+      "mean_token_accuracy": 0.8323236957192421,
+      "num_tokens": 293475.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6387806066939997,
+      "eval_loss": 0.5858062505722046,
+      "eval_mean_token_accuracy": 0.8344625789065694,
+      "eval_num_tokens": 293475.0,
+      "eval_runtime": 96.6891,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 120
+    },
+    {
+      "entropy": 0.6209714934229851,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.45989227294921875,
+      "learning_rate": 3.425640532392555e-05,
+      "loss": 0.5755774974822998,
+      "mean_token_accuracy": 0.8373909905552864,
+      "num_tokens": 341699.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6218485641618108,
+      "eval_loss": 0.5702213644981384,
+      "eval_mean_token_accuracy": 0.8379409902317579,
+      "eval_num_tokens": 341699.0,
+      "eval_runtime": 96.6937,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 140
+    },
+    {
+      "entropy": 0.6020509973168373,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.33028408885002136,
+      "learning_rate": 3.9185384507224195e-05,
+      "loss": 0.5572737693786621,
+      "mean_token_accuracy": 0.8403014644980431,
+      "num_tokens": 390139.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.5861590830392616,
+      "eval_loss": 0.5533568859100342,
+      "eval_mean_token_accuracy": 0.8414873508519904,
+      "eval_num_tokens": 390139.0,
+      "eval_runtime": 96.8046,
+      "eval_samples_per_second": 14.214,
+      "eval_steps_per_second": 1.777,
+      "step": 160
+    },
+    {
+      "entropy": 0.5785866163671016,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.38747045397758484,
+      "learning_rate": 4.411436369052284e-05,
+      "loss": 0.5444605350494385,
+      "mean_token_accuracy": 0.8429206758737564,
+      "num_tokens": 441038.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.5659288874892301,
+      "eval_loss": 0.5446343421936035,
+      "eval_mean_token_accuracy": 0.8433043364175531,
+      "eval_num_tokens": 441038.0,
+      "eval_runtime": 96.6663,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 180
+    },
+    {
+      "entropy": 0.5834021463990211,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.5842693448066711,
+      "learning_rate": 4.904334287382148e-05,
+      "loss": 0.5525710105895996,
+      "mean_token_accuracy": 0.8431421548128128,
+      "num_tokens": 492396.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.5828779492267343,
+      "eval_loss": 0.5379067659378052,
+      "eval_mean_token_accuracy": 0.8463593367920366,
+      "eval_num_tokens": 492396.0,
+      "eval_runtime": 96.639,
+      "eval_samples_per_second": 14.239,
+      "eval_steps_per_second": 1.78,
+      "step": 200
+    },
+    {
+      "entropy": 0.5798383548855781,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.32482433319091797,
+      "learning_rate": 5.397232205712012e-05,
+      "loss": 0.541356086730957,
+      "mean_token_accuracy": 0.8450109317898751,
+      "num_tokens": 543666.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.5595459116752758,
+      "eval_loss": 0.5338172316551208,
+      "eval_mean_token_accuracy": 0.8457894626744958,
+      "eval_num_tokens": 543666.0,
+      "eval_runtime": 96.6567,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 1.779,
+      "step": 220
+    },
+    {
+      "entropy": 0.5551654525101185,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.3377026915550232,
+      "learning_rate": 5.890130124041876e-05,
+      "loss": 0.5221311569213867,
+      "mean_token_accuracy": 0.8497206285595894,
+      "num_tokens": 591173.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.556636971789737,
+      "eval_loss": 0.5251970291137695,
+      "eval_mean_token_accuracy": 0.848015770316124,
+      "eval_num_tokens": 591173.0,
+      "eval_runtime": 96.7942,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 240
+    },
+    {
+      "entropy": 0.557022999972105,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.29898500442504883,
+      "learning_rate": 6.38302804237174e-05,
+      "loss": 0.5189533710479737,
+      "mean_token_accuracy": 0.8498866230249404,
+      "num_tokens": 640179.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.5586709416536397,
+      "eval_loss": 0.5241001844406128,
+      "eval_mean_token_accuracy": 0.8486643495947815,
+      "eval_num_tokens": 640179.0,
+      "eval_runtime": 96.8003,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 260
+    },
+    {
+      "entropy": 0.5592786997556687,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.3030458986759186,
+      "learning_rate": 6.875925960701604e-05,
+      "loss": 0.5240638732910157,
+      "mean_token_accuracy": 0.8491924479603767,
+      "num_tokens": 687183.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.566004884624204,
+      "eval_loss": 0.5177627205848694,
+      "eval_mean_token_accuracy": 0.8504269362189049,
+      "eval_num_tokens": 687183.0,
+      "eval_runtime": 96.7261,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 280
+    },
+    {
+      "entropy": 0.5667563803493977,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.33415117859840393,
+      "learning_rate": 7.368823879031469e-05,
+      "loss": 0.5269570350646973,
+      "mean_token_accuracy": 0.8470688253641129,
+      "num_tokens": 736370.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.5622092397060505,
+      "eval_loss": 0.5121559500694275,
+      "eval_mean_token_accuracy": 0.8510552939287451,
+      "eval_num_tokens": 736370.0,
+      "eval_runtime": 96.7226,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 1.778,
+      "step": 300
+    },
+    {
+      "entropy": 0.5403155937790871,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.294078528881073,
+      "learning_rate": 7.861721797361333e-05,
+      "loss": 0.5049179553985595,
+      "mean_token_accuracy": 0.8532818391919136,
+      "num_tokens": 785330.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.5470159045139025,
+      "eval_loss": 0.5063807368278503,
+      "eval_mean_token_accuracy": 0.8529217101113741,
+      "eval_num_tokens": 785330.0,
+      "eval_runtime": 96.7338,
+      "eval_samples_per_second": 14.225,
+      "eval_steps_per_second": 1.778,
+      "step": 320
+    },
+    {
+      "entropy": 0.5468257948756218,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.35112708806991577,
+      "learning_rate": 8.354619715691196e-05,
+      "loss": 0.5023184776306152,
+      "mean_token_accuracy": 0.8546695247292518,
+      "num_tokens": 831827.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.5112988160446633,
+      "eval_loss": 0.5058440566062927,
+      "eval_mean_token_accuracy": 0.8527335046335708,
+      "eval_num_tokens": 831827.0,
+      "eval_runtime": 96.7904,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 340
+    },
+    {
+      "entropy": 0.5497813194990158,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.29629987478256226,
+      "learning_rate": 8.84751763402106e-05,
+      "loss": 0.5108835220336914,
+      "mean_token_accuracy": 0.8519914656877517,
+      "num_tokens": 880341.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.5359154198058816,
+      "eval_loss": 0.5016657114028931,
+      "eval_mean_token_accuracy": 0.853807148891826,
+      "eval_num_tokens": 880341.0,
+      "eval_runtime": 96.7998,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 360
+    },
+    {
+      "entropy": 0.5497635789215565,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.27656999230384827,
+      "learning_rate": 9.340415552350925e-05,
+      "loss": 0.5135146141052246,
+      "mean_token_accuracy": 0.8511542648077011,
+      "num_tokens": 928613.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.5325134100262509,
+      "eval_loss": 0.5014682412147522,
+      "eval_mean_token_accuracy": 0.853295857130095,
+      "eval_num_tokens": 928613.0,
+      "eval_runtime": 96.7549,
+      "eval_samples_per_second": 14.222,
+      "eval_steps_per_second": 1.778,
+      "step": 380
+    },
+    {
+      "entropy": 0.5333064921200276,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.29332685470581055,
+      "learning_rate": 9.833313470680789e-05,
+      "loss": 0.4961216926574707,
+      "mean_token_accuracy": 0.854674169421196,
+      "num_tokens": 976385.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.5185546900990398,
+      "eval_loss": 0.49629414081573486,
+      "eval_mean_token_accuracy": 0.8552317674769911,
+      "eval_num_tokens": 976385.0,
+      "eval_runtime": 96.677,
+      "eval_samples_per_second": 14.233,
+      "eval_steps_per_second": 1.779,
+      "step": 400
+    },
+    {
+      "entropy": 0.512799974435415,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.3318716287612915,
+      "learning_rate": 9.906708466869441e-05,
+      "loss": 0.47603812217712405,
+      "mean_token_accuracy": 0.8589456249506046,
+      "num_tokens": 1023263.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5066671217250269,
+      "eval_loss": 0.4964885711669922,
+      "eval_mean_token_accuracy": 0.8553905123194983,
+      "eval_num_tokens": 1023263.0,
+      "eval_runtime": 96.7929,
+      "eval_samples_per_second": 14.216,
+      "eval_steps_per_second": 1.777,
+      "step": 420
+    },
+    {
+      "entropy": 0.5064023032784462,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.2704213857650757,
+      "learning_rate": 9.904691799913211e-05,
+      "loss": 0.4693995475769043,
+      "mean_token_accuracy": 0.8599134266376496,
+      "num_tokens": 1072585.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5176948705731437,
+      "eval_loss": 0.4949384331703186,
+      "eval_mean_token_accuracy": 0.8547424535418666,
+      "eval_num_tokens": 1072585.0,
+      "eval_runtime": 96.7659,
+      "eval_samples_per_second": 14.22,
+      "eval_steps_per_second": 1.777,
+      "step": 440
+    },
+    {
+      "entropy": 0.4914680339396,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.2915250360965729,
+      "learning_rate": 9.901181958461517e-05,
+      "loss": 0.45717477798461914,
+      "mean_token_accuracy": 0.8620341300964356,
+      "num_tokens": 1124634.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5079261915281762,
+      "eval_loss": 0.49107423424720764,
+      "eval_mean_token_accuracy": 0.8562383894310441,
+      "eval_num_tokens": 1124634.0,
+      "eval_runtime": 96.796,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 460
+    },
+    {
+      "entropy": 0.4995751313865185,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.26716962456703186,
+      "learning_rate": 9.896180001034665e-05,
+      "loss": 0.4605551719665527,
+      "mean_token_accuracy": 0.8633834689855575,
+      "num_tokens": 1173983.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5191026584700097,
+      "eval_loss": 0.4908095896244049,
+      "eval_mean_token_accuracy": 0.8568453407564829,
+      "eval_num_tokens": 1173983.0,
+      "eval_runtime": 96.7088,
+      "eval_samples_per_second": 14.228,
+      "eval_steps_per_second": 1.779,
+      "step": 480
+    },
+    {
+      "entropy": 0.5057077065110207,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.30032873153686523,
+      "learning_rate": 9.889687436154796e-05,
+      "loss": 0.46781325340270996,
+      "mean_token_accuracy": 0.8618202343583107,
+      "num_tokens": 1219697.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5059137819118278,
+      "eval_loss": 0.48817896842956543,
+      "eval_mean_token_accuracy": 0.8574246327544368,
+      "eval_num_tokens": 1219697.0,
+      "eval_runtime": 96.6929,
+      "eval_samples_per_second": 14.231,
+      "eval_steps_per_second": 1.779,
+      "step": 500
+    },
+    {
+      "entropy": 0.5053309261798858,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.27454873919487,
+      "learning_rate": 9.881706221890921e-05,
+      "loss": 0.467315149307251,
+      "mean_token_accuracy": 0.8609159409999847,
+      "num_tokens": 1269844.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.507903284458227,
+      "eval_loss": 0.4845135807991028,
+      "eval_mean_token_accuracy": 0.8583392351865768,
+      "eval_num_tokens": 1269844.0,
+      "eval_runtime": 96.7187,
+      "eval_samples_per_second": 14.227,
+      "eval_steps_per_second": 1.778,
+      "step": 520
+    },
+    {
+      "entropy": 0.5020621493458748,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.3254142105579376,
+      "learning_rate": 9.872238765268402e-05,
+      "loss": 0.4629322052001953,
+      "mean_token_accuracy": 0.8609199434518814,
+      "num_tokens": 1316433.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5141669095948685,
+      "eval_loss": 0.4836900532245636,
+      "eval_mean_token_accuracy": 0.8587847213412441,
+      "eval_num_tokens": 1316433.0,
+      "eval_runtime": 96.661,
+      "eval_samples_per_second": 14.235,
+      "eval_steps_per_second": 1.779,
+      "step": 540
+    },
+    {
+      "entropy": 0.49302729964256287,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.2609409689903259,
+      "learning_rate": 9.861287921543032e-05,
+      "loss": 0.4652863025665283,
+      "mean_token_accuracy": 0.862071031332016,
+      "num_tokens": 1363826.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.5069904422690702,
+      "eval_loss": 0.48222532868385315,
+      "eval_mean_token_accuracy": 0.8590792926938035,
+      "eval_num_tokens": 1363826.0,
+      "eval_runtime": 96.5653,
+      "eval_samples_per_second": 14.249,
+      "eval_steps_per_second": 1.781,
+      "step": 560
+    },
+    {
+      "entropy": 0.5024241611361504,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.3021513521671295,
+      "learning_rate": 9.848856993339911e-05,
+      "loss": 0.4593523979187012,
+      "mean_token_accuracy": 0.8627139970660209,
+      "num_tokens": 1412401.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.4920485499639844,
+      "eval_loss": 0.4806499183177948,
+      "eval_mean_token_accuracy": 0.859854532535686,
+      "eval_num_tokens": 1412401.0,
+      "eval_runtime": 96.8525,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 580
+    },
+    {
+      "entropy": 0.48968671932816504,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.2713676989078522,
+      "learning_rate": 9.834949729657438e-05,
+      "loss": 0.4635486602783203,
+      "mean_token_accuracy": 0.8626158565282822,
+      "num_tokens": 1460366.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5075812516517417,
+      "eval_loss": 0.48092299699783325,
+      "eval_mean_token_accuracy": 0.8594943108253701,
+      "eval_num_tokens": 1460366.0,
+      "eval_runtime": 96.8801,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 600
+    },
+    {
+      "entropy": 0.49362134337425234,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.26279115676879883,
+      "learning_rate": 9.819570324736658e-05,
+      "loss": 0.45868754386901855,
+      "mean_token_accuracy": 0.8646741047501564,
+      "num_tokens": 1511476.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.4934931300407232,
+      "eval_loss": 0.47820955514907837,
+      "eval_mean_token_accuracy": 0.8611027102830798,
+      "eval_num_tokens": 1511476.0,
+      "eval_runtime": 96.865,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 620
+    },
+    {
+      "entropy": 0.498298654705286,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.2743106484413147,
+      "learning_rate": 9.802723416796326e-05,
+      "loss": 0.4625910758972168,
+      "mean_token_accuracy": 0.8644399270415306,
+      "num_tokens": 1560069.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.49935419944136644,
+      "eval_loss": 0.4788094460964203,
+      "eval_mean_token_accuracy": 0.8595240989396739,
+      "eval_num_tokens": 1560069.0,
+      "eval_runtime": 96.8858,
+      "eval_samples_per_second": 14.202,
+      "eval_steps_per_second": 1.775,
+      "step": 640
+    },
+    {
+      "entropy": 0.5169856667518615,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.2438548356294632,
+      "learning_rate": 9.784414086634099e-05,
+      "loss": 0.4727236270904541,
+      "mean_token_accuracy": 0.8596569061279297,
+      "num_tokens": 1603766.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.48505116392706715,
+      "eval_loss": 0.4763250946998596,
+      "eval_mean_token_accuracy": 0.8609154699153678,
+      "eval_num_tokens": 1603766.0,
+      "eval_runtime": 96.8748,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 660
+    },
+    {
+      "entropy": 0.48837797343730927,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.2245505303144455,
+      "learning_rate": 9.764647856094226e-05,
+      "loss": 0.45554113388061523,
+      "mean_token_accuracy": 0.8651937782764435,
+      "num_tokens": 1654672.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.47867353970921317,
+      "eval_loss": 0.47178956866264343,
+      "eval_mean_token_accuracy": 0.8621448333873305,
+      "eval_num_tokens": 1654672.0,
+      "eval_runtime": 96.8536,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 1.776,
+      "step": 680
+    },
+    {
+      "entropy": 0.4763235367834568,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.24530072510242462,
+      "learning_rate": 9.743430686402244e-05,
+      "loss": 0.44700350761413576,
+      "mean_token_accuracy": 0.8670215323567391,
+      "num_tokens": 1706860.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5032368693933931,
+      "eval_loss": 0.466962605714798,
+      "eval_mean_token_accuracy": 0.8626160392927569,
+      "eval_num_tokens": 1706860.0,
+      "eval_runtime": 96.8466,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 700
+    },
+    {
+      "entropy": 0.48205163702368736,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.335552453994751,
+      "learning_rate": 9.720768976367141e-05,
+      "loss": 0.4490774631500244,
+      "mean_token_accuracy": 0.8643530890345573,
+      "num_tokens": 1754431.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.4913890836543815,
+      "eval_loss": 0.46917298436164856,
+      "eval_mean_token_accuracy": 0.862456685235334,
+      "eval_num_tokens": 1754431.0,
+      "eval_runtime": 96.875,
+      "eval_samples_per_second": 14.204,
+      "eval_steps_per_second": 1.775,
+      "step": 720
+    },
+    {
+      "entropy": 0.48629910722374914,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.2625334560871124,
+      "learning_rate": 9.696669560451573e-05,
+      "loss": 0.4475870132446289,
+      "mean_token_accuracy": 0.8631389886140823,
+      "num_tokens": 1804368.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.4911207033451213,
+      "eval_loss": 0.46702805161476135,
+      "eval_mean_token_accuracy": 0.8621654281782549,
+      "eval_num_tokens": 1804368.0,
+      "eval_runtime": 96.8685,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 740
+    },
+    {
+      "entropy": 0.475378530472517,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.2406812161207199,
+      "learning_rate": 9.671139706710683e-05,
+      "loss": 0.4375624179840088,
+      "mean_token_accuracy": 0.8660449728369712,
+      "num_tokens": 1850859.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.48983157044926356,
+      "eval_loss": 0.4654311239719391,
+      "eval_mean_token_accuracy": 0.8626760878535205,
+      "eval_num_tokens": 1850859.0,
+      "eval_runtime": 96.8581,
+      "eval_samples_per_second": 14.206,
+      "eval_steps_per_second": 1.776,
+      "step": 760
+    },
+    {
+      "entropy": 0.4725016601383686,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.2215867042541504,
+      "learning_rate": 9.644187114600152e-05,
+      "loss": 0.44096899032592773,
+      "mean_token_accuracy": 0.8674400866031646,
+      "num_tokens": 1901155.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.4720242848229963,
+      "eval_loss": 0.46539217233657837,
+      "eval_mean_token_accuracy": 0.8630018580791562,
+      "eval_num_tokens": 1901155.0,
+      "eval_runtime": 96.8222,
+      "eval_samples_per_second": 14.212,
+      "eval_steps_per_second": 1.776,
+      "step": 780
+    },
+    {
+      "entropy": 0.4846533551812172,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.2640039920806885,
+      "learning_rate": 9.615819912654142e-05,
+      "loss": 0.451138973236084,
+      "mean_token_accuracy": 0.8665898993611336,
+      "num_tokens": 1949864.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.49363225871740385,
+      "eval_loss": 0.463025838136673,
+      "eval_mean_token_accuracy": 0.8635496410519578,
+      "eval_num_tokens": 1949864.0,
+      "eval_runtime": 96.8241,
+      "eval_samples_per_second": 14.211,
+      "eval_steps_per_second": 1.776,
+      "step": 800
+    },
+    {
+      "entropy": 0.4358747288202628,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.28732356429100037,
+      "learning_rate": 9.586046656033853e-05,
+      "loss": 0.39547598361968994,
+      "mean_token_accuracy": 0.8815080141409849,
+      "num_tokens": 1998717.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.4406000122774479,
+      "eval_loss": 0.4686899781227112,
+      "eval_mean_token_accuracy": 0.8642892743958983,
+      "eval_num_tokens": 1998717.0,
+      "eval_runtime": 96.8701,
+      "eval_samples_per_second": 14.205,
+      "eval_steps_per_second": 1.776,
+      "step": 820
+    },
+    {
+      "entropy": 0.42045016661286355,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.2728928029537201,
+      "learning_rate": 9.554876323947397e-05,
+      "loss": 0.3843829154968262,
+      "mean_token_accuracy": 0.879941563308239,
+      "num_tokens": 2050953.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.4346749455083248,
+      "eval_loss": 0.4729752540588379,
+      "eval_mean_token_accuracy": 0.8633614282968433,
+      "eval_num_tokens": 2050953.0,
+      "eval_runtime": 96.8024,
+      "eval_samples_per_second": 14.215,
+      "eval_steps_per_second": 1.777,
+      "step": 840
+    },
+    {
+      "entropy": 0.4300770156085491,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.265896737575531,
+      "learning_rate": 9.522318316941786e-05,
+      "loss": 0.39586963653564455,
+      "mean_token_accuracy": 0.8787443146109581,
+      "num_tokens": 2096118.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4557679929705553,
+      "eval_loss": 0.47163790464401245,
+      "eval_mean_token_accuracy": 0.8636134530222693,
+      "eval_num_tokens": 2096118.0,
+      "eval_runtime": 96.8457,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 860
+    },
+    {
+      "entropy": 0.4276770144701004,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.2702751159667969,
+      "learning_rate": 9.488382454067867e-05,
+      "loss": 0.3841589689254761,
+      "mean_token_accuracy": 0.8800378635525703,
+      "num_tokens": 2146168.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4377675843100215,
+      "eval_loss": 0.47373291850090027,
+      "eval_mean_token_accuracy": 0.8632283134515896,
+      "eval_num_tokens": 2146168.0,
+      "eval_runtime": 96.8483,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 880
+    },
+    {
+      "entropy": 0.4314340800046921,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.27613386511802673,
+      "learning_rate": 9.453078969919021e-05,
+      "loss": 0.3940398931503296,
+      "mean_token_accuracy": 0.8791197642683983,
+      "num_tokens": 2195193.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4267724165043166,
+      "eval_loss": 0.4770275056362152,
+      "eval_mean_token_accuracy": 0.8623441904783249,
+      "eval_num_tokens": 2195193.0,
+      "eval_runtime": 96.846,
+      "eval_samples_per_second": 14.208,
+      "eval_steps_per_second": 1.776,
+      "step": 900
+    },
+    {
+      "entropy": 0.41646718233823776,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.34193891286849976,
+      "learning_rate": 9.416418511544553e-05,
+      "loss": 0.3823988676071167,
+      "mean_token_accuracy": 0.881490395963192,
+      "num_tokens": 2241455.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.45252677469059477,
+      "eval_loss": 0.4692145884037018,
+      "eval_mean_token_accuracy": 0.8639034258071766,
+      "eval_num_tokens": 2241455.0,
+      "eval_runtime": 96.7456,
+      "eval_samples_per_second": 14.223,
+      "eval_steps_per_second": 1.778,
+      "step": 920
+    },
+    {
+      "entropy": 0.43459259644150733,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.2719372808933258,
+      "learning_rate": 9.378412135238683e-05,
+      "loss": 0.39447042942047117,
+      "mean_token_accuracy": 0.8789274036884308,
+      "num_tokens": 2286048.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.4377528974829718,
+      "eval_loss": 0.47256559133529663,
+      "eval_mean_token_accuracy": 0.8633056709239649,
+      "eval_num_tokens": 2286048.0,
+      "eval_runtime": 96.9805,
+      "eval_samples_per_second": 14.188,
+      "eval_steps_per_second": 1.774,
+      "step": 940
+    },
+    {
+      "entropy": 0.4245278872549534,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.303096741437912,
+      "learning_rate": 9.33907130320613e-05,
+      "loss": 0.3913075923919678,
+      "mean_token_accuracy": 0.8795066639780998,
+      "num_tokens": 2336997.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.43375605102195297,
+      "eval_loss": 0.4710635840892792,
+      "eval_mean_token_accuracy": 0.86402241400508,
+      "eval_num_tokens": 2336997.0,
+      "eval_runtime": 96.8813,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 1.775,
+      "step": 960
+    },
+    {
+      "entropy": 0.4291906014084816,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.33950385451316833,
+      "learning_rate": 9.298407880105248e-05,
+      "loss": 0.3936340093612671,
+      "mean_token_accuracy": 0.8765407994389534,
+      "num_tokens": 2385830.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.45549828871045,
+      "eval_loss": 0.4683294892311096,
+      "eval_mean_token_accuracy": 0.8633980186179627,
+      "eval_num_tokens": 2385830.0,
+      "eval_runtime": 96.8992,
+      "eval_samples_per_second": 14.2,
+      "eval_steps_per_second": 1.775,
+      "step": 980
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.878945033940992e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}